x265 Documentation

Introduction

Increasing demand for high definition and ultra-high definition video, along with an increasing desire for video on demand has led to exponential growth in demand for bandwidth and storage requirements. These challenges can be met by the new High Efficiency Video Coding (HEVC) standard, also known as H.265. The x265 HEVC encoder project was launched by MulticoreWare in 2013, aiming to provide the most efficient, highest performance HEVC video encoder.

About HEVC

The High Efficiency Video Coding (HEVC) was developed by the ISO/IEC Moving Picture Experts Group (MPEG) and ITU-T Video Coding Experts Group (VCEG), through their Joint Collaborative Team on Video Coding (JCT-VC). HEVC is also known as ISO/IEC 23008-2 MPEG-H Part 2 and ITU-T H.265. HEVC provides superior video quality and up to twice the data compression as the previous standard (H.264/MPEG-4 AVC). HEVC can support 8K Ultra High Definition video, with a picture size up to 8192x4320 pixels.

About x265

The primary objective of x265 is to become the best H.265/HEVC encoder available anywhere, offering the highest compression efficiency and the highest performance on a wide variety of hardware platforms. The x265 encoder is available as an open source library, published under the GPLv2 license. It is also available under a commercial license, enabling commercial companies to utilize and distribute x265 in their solutions without being subject to the restrictions of the GPL license.

x265 is developed by MulticoreWare, leaders in high performance software solutions, with backing from leading video technology providers including Telestream and Doremi Labs (and other companies who want to remain anonymous at this time), and with contributions from open source developers. x265 leverages many of the outstanding video encoding features and optimizations from the x264 AVC encoder project.

The x265 software is available for free under the GNU GPL 2 license, from https://bitbucket.org/multicoreware/x265_git. For commercial companies that wish to distribute x265 without being subject to the open source requirements of the GPL 2 license, commercial licenses are available with competitive terms. Contact license @ x265.com to inquire about commercial license terms.

While x265 is primarily designed as a video encoder software library, a command-line executable is provided to facilitate testing and development. We expect x265 to be utilized in many leading video hardware and software products and services in the coming months.

Command Line Options

Note that unless an option is listed as CLI ONLY the option is also supported by x265_param_parse(). The CLI uses getopt to parse the command line options so the short or long versions may be used and the long options may be truncated to the shortest unambiguous abbreviation. Users of the API must pass x265_param_parse() the full option name.

Preset and tune have special implications. The API user must call x265_param_default_preset() with the preset and tune parameters they wish to use, prior to calling x265_param_parse() to set any additional fields. The CLI does this for the user implicitly, so all CLI options are applied after the user’s preset and tune choices, regardless of the order of the arguments on the command line.

If there is an extra command line argument (not an option or an option value) the CLI will treat it as the input filename. This effectively makes the --input specifier optional for the input file. If there are two extra arguments, the second is treated as the output bitstream filename, making --output also optional if the input filename was implied. This makes x265 in.y4m out.hevc a valid command line. If there are more than two extra arguments, the CLI will consider this an error and abort.

Generally, when an option expects a string value from a list of strings the user may specify the integer ordinal of the value they desire. ie: --log-level 3 is equivalent to --log-level debug.

Executable Options

--help, -h

Display help text

CLI ONLY

--version, -V

Display version details

CLI ONLY

Command line executable return codes:

0. encode successful
1. unable to parse command line
2. unable to open encoder
3. unable to generate stream headers
4. encoder abort

Logging/Statistic Options

--log-level <integer|string>

Controls the level of information displayed on the console. Debug level enables per-frame QP, metric, and bitrate logging. Full level enables hash and weight logging. -1 disables all logging, except certain fatal errors, and can be specified by the string “none”.

  1. error
  2. warning
  3. info (default)
  4. debug
  5. full
--no-progress

Disable periodic progress reports from the CLI

CLI ONLY

--csv <filename>

Write encoding statistics to a Comma Separated Values log file. Creates the file if it doesn’t already exist. If --csv-log-level is 0, it adds one line per run. If --csv-log-level is greater than 0, it writes one line per frame. Default none

The following statistics are available when --csv-log-level is greater than or equal to 1:

Encode Order The frame order in which the encoder encodes.

Type Slice type of the frame.

POC Picture Order Count - The display order of the frames.

QP Quantization Parameter decided for the frame.

Bits Number of bits consumed by the frame.

Scenecut 1 if the frame is a scenecut, 0 otherwise.

RateFactor Applicable only when CRF is enabled. The rate factor depends on the CRF given by the user. This is used to determine the QP so as to target a certain quality.

BufferFill Bits available for the next frame. Includes bits carried over from the current frame.

BufferFillFinal Buffer bits available after removing the frame out of CPB.

UnclippedBufferFillFinal Unclipped buffer bits available after removing the frame out of CPB only used for csv logging purpose.

Latency Latency in terms of number of frames between when the frame was given in and when the frame is given out.

PSNR Peak signal to noise ratio for Y, U and V planes.

SSIM A quality metric that denotes the structural similarity between frames.

Ref lists POC of references in lists 0 and 1 for the frame.

Several statistics about the encoded bitstream and encoder performance are available when --csv-log-level is greater than or equal to 2:

I/P cost ratio: The ratio between the cost when a frame is decided as an I frame to that when it is decided as a P frame as computed from the quarter-resolution frame in look-ahead. This, in combination with other parameters such as position of the frame in the GOP, is used to decide scene transitions.

Analysis statistics:

CU Statistics percentage of CU modes.

Distortion Average luma and chroma distortion. Calculated as SSE is done on fenc and recon(after quantization).

Psy Energy Average psy energy calculated as the sum of absolute difference between source and recon energy. Energy is measured by sa8d minus SAD.

Residual Energy Average residual energy. SSE is calculated on fenc and pred(before quantization).

Luma/Chroma Values minimum, maximum and average(averaged by area) luma and chroma values of source for each frame.

PU Statistics percentage of PU modes at each depth.

Performance statistics:

DecideWait ms number of milliseconds the frame encoder had to wait, since the previous frame was retrieved by the API thread, before a new frame has been given to it. This is the latency introduced by slicetype decisions (lookahead).

Row0Wait ms number of milliseconds since the frame encoder received a frame to encode before its first row of CTUs is allowed to begin compression. This is the latency introduced by reference frames making reconstructed and filtered rows available.

Wall time ms number of milliseconds between the first CTU being ready to be compressed and the entire frame being compressed and the output NALs being completed.

Ref Wait Wall ms number of milliseconds between the first reference row being available and the last reference row becoming available.

Total CTU time ms the total time (measured in milliseconds) spent by worker threads compressing and filtering CTUs for this frame.

Stall Time ms the number of milliseconds of the reported wall time that were spent with zero worker threads, aka all compression was completely stalled.

Total frame time Total time spent to encode the frame.

Avg WPP the average number of worker threads working on this frame, at any given time. This value is sampled at the completion of each CTU. This shows the effectiveness of Wavefront Parallel Processing.

Row Blocks the number of times a worker thread had to abandon the row of CTUs it was encoding because the row above it was not far enough ahead for the necessary reference data to be available. This is more of a problem for P frames where some blocks are much more expensive than others.

--csv-log-level <integer>

Controls the level of detail (and size) of –csv log files

  1. summary (default)
  2. frame level logging
  3. frame level logging with performance statistics
--ssim, --no-ssim

Calculate and report Structural Similarity values. It is recommended to use --tune ssim if you are measuring ssim, else the results should not be used for comparison purposes. Default disabled

--psnr, --no-psnr

Calculate and report Peak Signal to Noise Ratio. It is recommended to use --tune psnr if you are measuring PSNR, else the results should not be used for comparison purposes. Default disabled

Performance Options

--asm <integer:false:string>, --no-asm

x265 will use all detected CPU SIMD architectures by default. You can disable all assembly by using --no-asm or you can specify a comma separated list of SIMD architectures to use, matching these strings: MMX2, SSE, SSE2, SSE3, SSSE3, SSE4, SSE4.1, SSE4.2, AVX, XOP, FMA4, AVX2, FMA3

Some higher architectures imply lower ones being present, this is handled implicitly.

One may also directly supply the CPU capability bitmap as an integer.

Note that by specifying this option you are overriding x265’s CPU detection and it is possible to do this wrong. You can cause encoder crashes by specifying SIMD architectures which are not supported on your CPU.

Default: auto-detected SIMD architectures

--frame-threads, -F <integer>

Number of concurrently encoded frames. Using a single frame thread gives a slight improvement in compression, since the entire reference frames are always available for motion compensation, but it has severe performance implications. Default is an autodetected count based on the number of CPU cores and whether WPP is enabled or not.

Over-allocation of frame threads will not improve performance, it will generally just increase memory use.

Values: any value between 0 and 16. Default is 0, auto-detect

--pools <string>, --numa-pools <string>

Comma separated list of threads per NUMA node. If “none”, then no worker pools are created and only frame parallelism is possible. If NULL or “” (default) x265 will use all available threads on each NUMA node:

'+'  is a special value indicating all cores detected on the node
'*'  is a special value indicating all cores detected on the node and all remaining nodes
'-'  is a special value indicating no cores on the node, same as '0'

example strings for a 4-node system:

""        - default, unspecified, all numa nodes are used for thread pools
"*"       - same as default
"none"    - no thread pools are created, only frame parallelism possible
"-"       - same as "none"
"10"      - allocate one pool, using up to 10 cores on all available nodes
"-,+"     - allocate one pool, using all cores on node 1
"+,-,+"   - allocate one pool, using only cores on nodes 0 and 2
"+,-,+,-" - allocate one pool, using only cores on nodes 0 and 2
"-,*"     - allocate one pool, using all cores on nodes 1, 2 and 3
"8,8,8,8" - allocate four pools with up to 8 threads in each pool
"8,+,+,+" - allocate two pools, the first with 8 threads on node 0, and the second with all cores on node 1,2,3

A thread pool dedicated to a given NUMA node is enabled only when the number of threads to be created on that NUMA node is explicitly mentioned in that corresponding position with the –pools option. Else, all threads are spawned from a single pool. The total number of threads will be determined by the number of threads assigned to the enabled NUMA nodes for that pool. The worker threads are be given affinity to all the enabled NUMA nodes for that pool and may migrate between them, unless explicitly specified as described above.

In the case that any threadpool has more than 64 threads, the threadpool may be broken down into multiple pools of 64 threads each; on 32-bit machines, this number is 32. All pools are given affinity to the NUMA nodes on which the original pool had affinity. For performance reasons, the last thread pool is spawned only if it has more than 32 threads for 64-bit machines, or 16 for 32-bit machines. If the total number of threads in the system doesn’t obey this constraint, we may spawn fewer threads than cores which has been empirically shown to be better for performance.

If the four pool features: --wpp, --pmode, --pme and --lookahead-slices are all disabled, then --pools is ignored and no thread pools are created.

If “none” is specified, then all four of the thread pool features are implicitly disabled.

Frame encoders are distributed between the available thread pools, and the encoder will never generate more thread pools than --frame-threads. The pools are used for WPP and for distributed analysis and motion search.

On Windows, the native APIs offer sufficient functionality to discover the NUMA topology and enforce the thread affinity that libx265 needs (so long as you have not chosen to target XP or Vista), but on POSIX systems it relies on libnuma for this functionality. If your target POSIX system is single socket, then building without libnuma is a perfectly reasonable option, as it will have no effect on the runtime behavior. On a multiple-socket system, a POSIX build of libx265 without libnuma will be less work efficient. See thread pools for more detail.

Default “”, one pool is created across all available NUMA nodes, with one thread allocated per detected hardware thread (logical CPU cores). In the case that the total number of threads is more than the maximum size that ATOMIC operations can handle (32 for 32-bit compiles, and 64 for 64-bit compiles), multiple thread pools may be spawned subject to the performance constraint described above.

Note that the string value will need to be escaped or quoted to protect against shell expansion on many platforms

--wpp, --no-wpp

Enable Wavefront Parallel Processing. The encoder may begin encoding a row as soon as the row above it is at least two CTUs ahead in the encode process. This gives a 3-5x gain in parallelism for about 1% overhead in compression efficiency.

This feature is implicitly disabled when no thread pool is present.

Default: Enabled

--pmode, --no-pmode

Parallel mode decision, or distributed mode analysis. When enabled the encoder will distribute the analysis work of each CU (merge, inter, intra) across multiple worker threads. Only recommended if x265 is not already saturating the CPU cores. In RD levels 3 and 4 it will be most effective if –rect is enabled. At RD levels 5 and 6 there is generally always enough work to distribute to warrant the overhead, assuming your CPUs are not already saturated.

–pmode will increase utilization without reducing compression efficiency. In fact, since the modes are all measured in parallel it makes certain early-outs impractical and thus you usually get slightly better compression when it is enabled (at the expense of not skipping improbable modes). This bypassing of early-outs can cause pmode to slow down encodes, especially at faster presets.

This feature is implicitly disabled when no thread pool is present.

Default disabled

--pme, --no-pme

Parallel motion estimation. When enabled the encoder will distribute motion estimation across multiple worker threads when more than two references require motion searches for a given CU. Only recommended if x265 is not already saturating CPU cores. --pmode is much more effective than this option, since the amount of work it distributes is substantially higher. With –pme it is not unusual for the overhead of distributing the work to outweigh the parallelism benefits.

This feature is implicitly disabled when no thread pool is present.

–pme will increase utilization on many core systems with no effect on the output bitstream.

Default disabled

--preset, -p <integer|string>

Sets parameters to preselected values, trading off compression efficiency against encoding speed. These parameters are applied before all other input parameters are applied, and so you can override any parameters that these values control. See presets for more detail.

  1. ultrafast
  2. superfast
  3. veryfast
  4. faster
  5. fast
  6. medium (default)
  7. slow
  8. slower
  9. veryslow
  10. placebo
--tune, -t <string>

Tune the settings for a particular type of source or situation. The changes will be applied after --preset but before all other parameters. Default none. See tunings for more detail.

Values: psnr, ssim, grain, zero-latency, fast-decode, animation.

--slices <integer>

Encode each incoming frame as multiple parallel slices that may be decoded independently. Support available only for rectangular slices that cover the entire width of the image.

Recommended for improving encoder performance only if frame-parallelism and WPP are unable to maximize utilization on given hardware.

Default: 1 slice per frame. Experimental feature

--copy-pic, --no-copy-pic

Allow encoder to copy input x265 pictures to internal frame buffers. When disabled, x265 will not make an internal copy of the input picture and will work with the application’s buffers. While this allows for deeper integration, it is the responsibility of the application to (a) ensure that the allocated picture has extra space for padding that will be done by the library, and (b) the buffers aren’t recycled until the library has completed encoding this frame (which can be figured out by tracking NALs output by x265)

Default: enabled

Input/Output File Options

These options all describe the input video sequence or, in the case of --dither, operations that are performed on the sequence prior to encode. All options dealing with files (names, formats, offsets or frame counts) are only applicable to the CLI application.

--input <filename>

Input filename, only raw YUV or Y4M supported. Use single dash for stdin. This option name will be implied for the first “extra” command line argument.

CLI ONLY

--y4m

Parse input stream as YUV4MPEG2 regardless of file extension, primarily intended for use with stdin (ie: --input - --y4m). This option is implied if the input filename has a “.y4m” extension

CLI ONLY

--input-depth <integer>

YUV only: Bit-depth of input file or stream

Values: any value between 8 and 16. Default is internal depth.

CLI ONLY

--frames <integer>

The number of frames intended to be encoded. It may be left unspecified, but when it is specified rate control can make use of this information. It is also used to determine if an encode is actually a stillpicture profile encode (single frame)

--dither

Enable high quality downscaling to the encoder’s internal bitdepth. Dithering is based on the diffusion of errors from one row of pixels to the next row of pixels in a picture. Only applicable when the input bit depth is larger than 8bits. Default disabled

CLI ONLY

--input-res <wxh>

YUV only: Source picture size [w x h]

CLI ONLY

--input-csp <integer|string>

Chroma Subsampling (YUV only): Only 4:0:0(monochrome), 4:2:0, 4:2:2, and 4:4:4 are supported at this time. The chroma subsampling format of your input must match your desired output chroma subsampling format (libx265 will not perform any chroma subsampling conversion), and it must be supported by the HEVC profile you have specified.

  1. i400 (4:0:0 monochrome) - Not supported by Main or Main10 profiles
  2. i420 (4:2:0 default) - Supported by all HEVC profiles
  3. i422 (4:2:2) - Not supported by Main, Main10 and Main12 profiles
  4. i444 (4:4:4) - Supported by Main 4:4:4, Main 4:4:4 10, Main 4:4:4 12, Main 4:4:4 16 Intra profiles
  5. nv12
  6. nv16
--fps <integer|float|numerator/denominator>

YUV only: Source frame rate

Range of values: positive int or float, or num/denom

--interlace <false|tff|bff>, --no-interlace
  1. progressive pictures (default)
  2. top field first
  3. bottom field first

HEVC encodes interlaced content as fields. Fields must be provided to the encoder in the correct temporal order. The source dimensions must be field dimensions and the FPS must be in units of fields per second. The decoder must re-combine the fields in their correct orientation for display.

--frame-dup, --no-frame-dup

Enable Adaptive Frame duplication. Replaces 2-3 near-identical frames with one frame and sets pic_struct based on frame doubling / tripling. Default disabled.

--dup-threshold <integer>

Frame similarity threshold can vary between 1 and 99. This requires Adaptive Frame Duplication to be enabled. Default 70.

--seek <integer>

Number of frames to skip at start of input file. Default 0

CLI ONLY

--frames, -f <integer>

Number of frames of input sequence to be encoded. Default 0 (all)

CLI ONLY

--output, -o <filename>

Bitstream output file name. If there are two extra CLI options, the first is implicitly the input filename and the second is the output filename, making the --output option optional.

The output file will always contain a raw HEVC bitstream, the CLI does not support any container file formats.

CLI ONLY

--output-depth, -D 8|10|12

Bitdepth of output HEVC bitstream, which is also the internal bit depth of the encoder. If the requested bit depth is not the bit depth of the linked libx265, it will attempt to bind libx265_main for an 8bit encoder, libx265_main10 for a 10bit encoder, or libx265_main12 for a 12bit encoder, with the same API version as the linked libx265.

If the output depth is not specified but --profile is specified, the output depth will be derived from the profile name.

CLI ONLY

--chunk-start <integer>

First frame of the chunk. Frames preceding this in display order will be encoded, however, they will be discarded in the bitstream. This feature can be enabled only in closed GOP structures. Default 0 (disabled).

--chunk-end <integer>

Last frame of the chunk. Frames following this in display order will be used in taking lookahead decisions, but they will not be encoded. This feature can be enabled only in closed GOP structures. Default 0 (disabled).

--field, --no-field

Enable or disable field coding. Default disabled.

Profile, Level, Tier

--profile, -P <string>

Enforce the requirements of the specified profile, ensuring the output stream will be decodable by a decoder which supports that profile. May abort the encode if the specified profile is impossible to be supported by the compile options chosen for the encoder (a high bit depth encoder will be unable to output bitstreams compliant with Main or MainStillPicture).

The following profiles are supported in x265.

8bit profiles:

* main, main-intra, mainstillpicture (or msp for short)
* main444-8, main444-intra, main444-stillpicture

See note below on signaling intra and stillpicture profiles.

10bit profiles:

* main10, main10-intra
* main422-10, main422-10-intra
* main444-10, main444-10-intra

12bit profiles:

* main12, main12-intra
* main422-12, main422-12-intra
* main444-12, main444-12-intra

CLI ONLY

API users must call x265_param_apply_profile() after configuring their param structure. Any changes made to the param structure after this call might make the encode non-compliant.

The CLI application will derive the output bit depth from the profile name if --output-depth is not specified.

--level-idc <integer|float>

Minimum decoder requirement level. Defaults to 0, which implies auto-detection by the encoder. If specified, the encoder will attempt to bring the encode specifications within that specified level. If the encoder is unable to reach the level it issues a warning and aborts the encode. If the requested requirement level is higher than the actual level, the actual requirement level is signaled.

Beware, specifying a decoder level will force the encoder to enable VBV for constant rate factor encodes, which may introduce non-determinism.

The value is specified as a float or as an integer with the level times 10, for example level 5.1 is specified as “5.1” or “51”, and level 5.0 is specified as “5.0” or “50”.

Annex A levels: 1, 2, 2.1, 3, 3.1, 4, 4.1, 5, 5.1, 5.2, 6, 6.1, 6.2, 8.5

--high-tier, --no-high-tier

If --level-idc has been specified, –high-tier allows the support of high tier at that level. The encoder will first attempt to encode at the specified level, main tier first, turning on high tier only if necessary and available at that level. If your requested level does not support a High tier, high tier will not be supported. If –no-high-tier has been specified, then the encoder will attempt to encode only at the main tier.

Default: enabled

--ref <1..16>

Max number of L0 references to be allowed. This number has a linear multiplier effect on the amount of work performed in motion search but will generally have a beneficial effect on compression and distortion.

Note that x265 allows up to 16 L0 references but the HEVC specification only allows a maximum of 8 total reference frames. So if you have B frames enabled only 7 L0 refs are valid and if you have --b-pyramid enabled (which is enabled by default in all presets), then only 6 L0 refs are the maximum allowed by the HEVC specification. If x265 detects that the total reference count is greater than 8, it will issue a warning that the resulting stream is non-compliant and it signals the stream as profile NONE and level NONE and will abort the encode unless --allow-non-conformance it specified. Compliant HEVC decoders may refuse to decode such streams.

Default 3

--allow-non-conformance, --no-allow-non-conformance

Allow libx265 to generate a bitstream with profile and level NONE. By default, it will abort any encode which does not meet strict level compliance. The two most likely causes for non-conformance are --ctu being too small, --ref being too high, or the bitrate or resolution being out of specification.

Default: disabled

--uhd-bd

Enable Ultra HD Blu-ray format support. If specified with incompatible encoding options, the encoder will attempt to modify/set the right encode specifications. If the encoder is unable to do so, this option will be turned OFF. Highly experimental.

Default: disabled

Note

--profile, --level-idc, and --high-tier are only intended for use when you are targeting a particular decoder (or decoders) with fixed resource limitations and must constrain the bitstream within those limits. Specifying a profile or level may lower the encode quality parameters to meet those requirements but it will never raise them. It may enable VBV constraints on a CRF encode.

Also note that x265 determines the decoder requirement profile and level in three steps. First, the user configures an x265_param structure with their suggested encoder options and then optionally calls x265_param_apply_profile() to enforce a specific profile (main, main10, etc). Second, an encoder is created from this x265_param instance and the --level-idc and --high-tier parameters are used to reduce bitrate or other features in order to enforce the target level. Finally, the encoder re-examines the final set of parameters and detects the actual minimum decoder requirement level and this is what is signaled in the bitstream headers. The detected decoder level will only use High tier if the user specified a High tier level.

The signaled profile will be determined by the encoder’s internal bitdepth and input color space. If --keyint is 0 or 1, then an intra variant of the profile will be signaled.

If --total-frames is 1, then a stillpicture variant will be signaled, but this parameter is not always set by applications, particularly not when the CLI uses stdin streaming or when libx265 is used by third-party applications.

Mode decision / Analysis

--rd <1..6>

Level of RDO in mode decision. The higher the value, the more exhaustive the analysis and the more rate distortion optimization is used. The lower the value the faster the encode, the higher the value the smaller the bitstream (in general). Default 3

Note that this table aims for accuracy but is not necessarily our final target behavior for each mode.

Level Description
0 sa8d mode and split decisions, intra w/ source pixels, currently not supported
1 recon generated (better intra), RDO merge/skip selection
2 RDO splits and merge/skip selection
3 RDO mode and split decisions, chroma residual used for sa8d
4 Currently same as 3
5 Adds RDO prediction decisions
6 Currently same as 5

Range of values: 1: least .. 6: full RDO analysis

Options which affect the coding unit quad-tree, sometimes referred to as the prediction quad-tree.

--ctu, -s <64|32|16>

Maximum CU size (width and height). The larger the maximum CU size, the more efficiently x265 can encode flat areas of the picture, giving large reductions in bitrate. However, this comes at a loss of parallelism with fewer rows of CUs that can be encoded in parallel, and less frame parallelism as well. Because of this the faster presets use a CU size of 32. Default: 64

--min-cu-size <32|16|8>

Minimum CU size (width and height). By using 16 or 32 the encoder will not analyze the cost of CUs below that minimum threshold, saving considerable amounts of compute with a predictable increase in bitrate. This setting has a large effect on performance on the faster presets.

Default: 8 (minimum 8x8 CU for HEVC, best compression efficiency)

Note

All encoders within a single process must use the same settings for the CU size range. --ctu and --min-cu-size must be consistent for all of them since the encoder configures several key global data structures based on this range.

--limit-refs <0|1|2|3>

When set to X265_REF_LIMIT_DEPTH (1) x265 will limit the references analyzed at the current depth based on the references used to code the 4 sub-blocks at the next depth. For example, a 16x16 CU will only use the references used to code its four 8x8 CUs.

When set to X265_REF_LIMIT_CU (2), the rectangular and asymmetrical partitions will only use references selected by the 2Nx2N motion search (including at the lowest depth which is otherwise unaffected by the depth limit).

When set to 3 (X265_REF_LIMIT_DEPTH && X265_REF_LIMIT_CU), the 2Nx2N motion search at each depth will only use references from the split CUs and the rect/amp motion searches at that depth will only use the reference(s) selected by 2Nx2N.

For all non-zero values of limit-refs, the current depth will evaluate intra mode (in inter slices), only if intra mode was chosen as the best mode for at least one of the 4 sub-blocks.

You can often increase the number of references you are using (within your decoder level limits) if you enable one or both of these flags.

Default 3.

--limit-modes, --no-limit-modes

When enabled, limit-modes will limit modes analyzed for each CU using cost metrics from the 4 sub-CUs. When multiple inter modes like --rect and/or --amp are enabled, this feature will use motion cost heuristics from the 4 sub-CUs to bypass modes that are unlikely to be the best choice. This can significantly improve performance when rect and/or --amp are enabled at minimal compression efficiency loss.

--rect, --no-rect

Enable analysis of rectangular motion partitions Nx2N and 2NxN (50/50 splits, two directions). Default disabled

--amp, --no-amp

Enable analysis of asymmetric motion partitions (75/25 splits, four directions). At RD levels 0 through 4, AMP partitions are only considered at CU sizes 32x32 and below. At RD levels 5 and 6, it will only consider AMP partitions as merge candidates (no motion search) at 64x64, and as merge or inter candidates below 64x64.

The AMP partitions which are searched are derived from the current best inter partition. If Nx2N (vertical rectangular) is the best current prediction, then left and right asymmetrical splits will be evaluated. If 2NxN (horizontal rectangular) is the best current prediction, then top and bottom asymmetrical splits will be evaluated, If 2Nx2N is the best prediction, and the block is not a merge/skip, then all four AMP partitions are evaluated.

This setting has no effect if rectangular partitions are disabled. Default disabled

--early-skip, --no-early-skip

Measure 2Nx2N merge candidates first; if no residual is found, additional modes at that depth are not analysed. Default disabled

--rskip <0|1|2>

This option determines early exit from CU depth recursion in modes 1 and 2. When a skip CU is found, additional heuristics (depending on the RD level and rskip mode) are used to decide whether to terminate recursion. The following table summarizes the behavior.

RD Level Rskip Mode Skip Recursion Heuristic
0 - 4 1 Neighbour costs and CU homogenity.
5 - 6 1 Comparison with inter2Nx2N.
0 - 6 2 CU edge density.

Provides minimal quality degradation at good performance gains for non-zero modes. --rskip mode 0 means disabled. Default: 1, disabled when --tune grain is used.

--rskip-edge-threshold <0..100>

Denotes the minimum expected edge-density percentage within the CU, below which the recursion is skipped. Internally normalized to decimal value in x265 library. Recommended low thresholds for slow encodes and high for fast encodes. Default: 5, requires --rskip mode 2 to be enabled.

--splitrd-skip, --no-splitrd-skip

Enable skipping split RD analysis when sum of split CU rdCost larger than one split CU rdCost for Intra CU. Default disabled.

--fast-intra, --no-fast-intra

Perform an initial scan of every fifth intra angular mode, then check modes +/- 2 distance from the best mode, then +/- 1 distance from the best mode, effectively performing a gradient descent. When enabled 10 modes in total are checked. When disabled all 33 angular modes are checked. Only applicable for --rd levels 4 and below (medium preset and faster).

--b-intra, --no-b-intra

Enables the evaluation of intra modes in B slices. Default disabled.

--cu-lossless, --no-cu-lossless

For each CU, evaluate lossless (transform and quant bypass) encode of the best non-lossless mode option as a potential rate distortion optimization. If the global option --lossless has been specified, all CUs will be encoded as lossless unconditionally regardless of whether this option was enabled. Default disabled.

Only effective at RD levels 3 and above, which perform RDO mode decisions.

--tskip-fast, --no-tskip-fast

Only evaluate transform skip for NxN intra predictions (4x4 blocks). Only applicable if transform skip is enabled. For chroma, only evaluate if luma used tskip. Inter block tskip analysis is unmodified. Default disabled

--rd-refine, --no-rd-refine

For each analysed CU, calculate R-D cost on the best partition mode for a range of QP values, to find the optimal rounding effect. Default disabled.

Only effective at RD levels 5 and 6

Analysis re-use options, to improve performance when encoding the same sequence multiple times (presumably at varying bitrates). The encoder will not reuse analysis if slice type parameters do not match.

--analysis-save <filename>

Encoder outputs analysis information of each frame. Analysis data from save mode is written to the file specified. Requires cutree, pmode to be off. Default disabled.

The amount of analysis data stored is determined by --analysis-save-reuse-level.

--analysis-load <filename>

Encoder reuses analysis information from the file specified. By reading the analysis data written by an earlier encode of the same sequence, substantial redundant work may be avoided. Requires cutree, pmode to be off. Default disabled.

The amount of analysis data reused is determined by --analysis-load-reuse-level.

--analysis-reuse-file <filename>

Specify a filename for --multi-pass-opt-analysis and option:–multi-pass-opt-distortion. If no filename is specified, x265_analysis.dat is used.

--analysis-save-reuse-level <1..10>, --analysis-load-reuse-level <1..10>

‘analysis-save-reuse-level’ denotes the amount of information stored during --analysis-save and ‘analysis-load-reuse-level’ denotes the amount of information reused during --analysis-load. Higher the value, higher the information stored/reused, faster the encode. Default 0. If not set during analysis-save/load, the encoder will internally configure them to 5.

Note that --analysis-save-reuse-level and --analysis-load-reuse-level must be paired with --analysis-save and --analysis-load respectively.

Level Description
1 Lookahead information
2 to 4 Level 1 + intra/inter modes, ref’s
5 and 6 Level 2 + rect-amp
7 Level 5 + AVC size CU refinement
8 and 9 Level 5 + AVC size Full CU analysis-info
10 Level 5 + Full CU analysis-info
--refine-mv-type <string>

Reuse MV information received through API call. Currently receives information for AVC size and the accepted string input is “avc”. Default is disabled.

--refine-ctu-distortion <0/1>

Store/normalize ctu distortion in analysis-save/load. 0 - Disabled. 1 - Save ctu distortion to the analysis file specified during --analysis-save.

Load CTU distortion from the analysis file and normalize it across every frame during --analysis-load.

Default 0.

--scale-factor

Factor by which input video is scaled down for analysis save mode. This option should be coupled with --analysis-load/--analysis-save at reuse levels 1 to 6 and 10. The ctu size of load can either be the same as that of save or double the size of save. Default 0.

--refine-intra <0..4>

Enables refinement of intra blocks in current encode.

Level 0 - Forces both mode and depth from the save encode.

Level 1 - Evaluates all intra modes at current depth(n) and at depth (n+1) when current block size is one greater than the min-cu-size. Forces modes for larger blocks.

Level 2 - In addition to the functionality of level 1, at all depths, force (a) only depth when angular mode is chosen by the save encode. (b) depth and mode when other intra modes are chosen by the save encode.

Level 3 - Perform analysis of intra modes for depth reused from first encode.

Level 4 - Does not reuse any analysis information - redo analysis for the intra block.

Default 0.

--refine-inter <0..3>

Enables refinement of inter blocks in current encode.

Level 0 - Forces both mode and depth from the save encode.

Level 1 - Evaluates all inter modes at current depth(n) and at depth (n+1) when current block size is one greater than the min-cu-size. Forces modes for larger blocks.

Level 2 - In addition to the functionality of level 1, restricts the modes evaluated when specific modes are decided as the best mode by the save encode.

2nx2n in save encode - disable re-evaluation of rect and amp.

skip in save encode - re-evaluates only skip, merge and 2nx2n modes.

Level 3 - Perform analysis of inter modes while reusing depths from the save encode.

Default 0.

--dynamic-refine, --no-dynamic-refine

Dynamically switches --refine-inter levels 0-3 based on the content and the encoder settings. It is recommended to use --refine-intra 4 with dynamic refinement. Default disabled.

--refine-mv <1..3>

Enables refinement of motion vector for scaled video. Evaluates the best motion vector based on the level selected. Default 1.

Level 1 - Search around scaled MV.

Level 2 - Level 1 + Search around best AMVP cand.

Level 3 - Level 2 + Search around the other AMVP cand.

Options which affect the transform unit quad-tree, sometimes referred to as the residual quad-tree (RQT).

--rdoq-level <0|1|2>, --no-rdoq-level

Specify the amount of rate-distortion analysis to use within quantization:

At level 0 rate-distortion cost is not considered in quant

At level 1 rate-distortion cost is used to find optimal rounding values for each level (and allows psy-rdoq to be effective). It trades-off the signaling cost of the coefficient vs its post-inverse quant distortion from the pre-quant coefficient. When --psy-rdoq is enabled, this formula is biased in favor of more energy in the residual (larger coefficient absolute levels)

At level 2 rate-distortion cost is used to make decimate decisions on each 4x4 coding group, including the cost of signaling the group within the group bitmap. If the total distortion of not signaling the entire coding group is less than the rate cost, the block is decimated. Next, it applies rate-distortion cost analysis to the last non-zero coefficient, which can result in many (or all) of the coding groups being decimated. Psy-rdoq is less effective at preserving energy when RDOQ is at level 2, since it only has influence over the level distortion costs.

--tu-intra-depth <1..4>

The transform unit (residual) quad-tree begins with the same depth as the coding unit quad-tree, but the encoder may decide to further split the transform unit tree if it improves compression efficiency. This setting limits the number of extra recursion depth which can be attempted for intra coded units. Default: 1, which means the residual quad-tree is always at the same depth as the coded unit quad-tree

Note that when the CU intra prediction is NxN (only possible with 8x8 CUs), a TU split is implied, and thus the residual quad-tree begins at 4x4 and cannot split any further.

--tu-inter-depth <1..4>

The transform unit (residual) quad-tree begins with the same depth as the coding unit quad-tree, but the encoder may decide to further split the transform unit tree if it improves compression efficiency. This setting limits the number of extra recursion depth which can be attempted for inter coded units. Default: 1. which means the residual quad-tree is always at the same depth as the coded unit quad-tree unless the CU was coded with rectangular or AMP partitions, in which case a TU split is implied and thus the residual quad-tree begins one layer below the CU quad-tree.

--limit-tu <0..4>

Enables early exit from TU depth recursion, for inter coded blocks.

Level 1 - decides to recurse to next higher depth based on cost comparison of full-size TU and split TU.

Level 2 - based on first split subTU’s depth, limits recursion of other split subTUs.

Level 3 - based on the average depth of the co-located and the neighbor CUs’ TU depth, limits recursion of the current CU.

Level 4 - uses the depth of the neighboring/ co-located CUs TU depth to limit the 1st subTU depth. The 1st subTU depth is taken as the limiting depth for the other subTUs.

Enabling levels 3 or 4 may cause a mismatch in the output bitstreams between --analysis-save and --analysis-load as all neighboring CUs TU depth may not be available in the --analysis-load run as only the best mode’s information is available to it.

Default: 0

--nr-intra <integer>, --nr-inter <integer>

Noise reduction - an adaptive deadzone applied after DCT (subtracting from DCT coefficients), before quantization. It does no pixel-level filtering, doesn’t cross DCT block boundaries, has no overlap, The higher the strength value parameter, the more aggressively it will reduce noise.

Enabling noise reduction will make outputs diverge between different numbers of frame threads. Outputs will be deterministic but the outputs of -F2 will no longer match the outputs of -F3, etc.

Values: any value in range of 0 to 2000. Default 0 (disabled).

--tskip, --no-tskip

Enable evaluation of transform skip (bypass DCT but still use quantization) coding for 4x4 TU coded blocks.

Only effective at RD levels 3 and above, which perform RDO mode decisions. Default disabled

--rdpenalty <0..2>

When set to 1, transform units of size 32x32 are given a 4x bit cost penalty compared to smaller transform units, in intra coded CUs in P or B slices.

When set to 2, transform units of size 32x32 are not even attempted, unless otherwise required by the maximum recursion depth. For this option to be effective with 32x32 intra CUs, --tu-intra-depth must be at least 2. For it to be effective with 64x64 intra CUs, --tu-intra-depth must be at least 3.

Note that in HEVC an intra transform unit (a block of the residual quad-tree) is also a prediction unit, meaning that the intra prediction signal is generated for each TU block, the residual subtracted and then coded. The coding unit simply provides the prediction modes that will be used when predicting all of the transform units within the CU. This means that when you prevent 32x32 intra transform units, you are preventing 32x32 intra predictions.

Default 0, disabled.

Values: 0:disabled 1:4x cost penalty 2:force splits

--max-tu-size <32|16|8|4>

Maximum TU size (width and height). The residual can be more efficiently compressed by the DCT transform when the max TU size is larger, but at the expense of more computation. Transform unit quad-tree begins at the same depth of the coded tree unit, but if the maximum TU size is smaller than the CU size then transform QT begins at the depth of the max-tu-size. Default: 32.

--dynamic-rd <0..4>

Increases the RD level at points where quality drops due to VBV rate control enforcement. The number of CUs for which the RD is reconfigured is determined based on the strength. Strength 1 gives the best FPS, strength 4 gives the best SSIM. Strength 0 switches this feature off. Default: 0.

Effective for RD levels 4 and below.

--ssim-rd, --no-ssim-rd

Enable/Disable SSIM RDO. SSIM is a better perceptual quality assessment method as compared to MSE. SSIM based RDO calculation is based on residual divisive normalization scheme. This normalization is consistent with the luminance and contrast masking effect of Human Visual System. It is used for mode selection during analysis of CTUs and can achieve significant gain in terms of objective quality metrics SSIM and PSNR. It only has effect on presets which use RDO-based mode decisions (--rd 3 and above).

Temporal / motion search options

--max-merge <1..5>

Maximum number of neighbor (spatial and temporal) candidate blocks that the encoder may consider for merging motion predictions. If a merge candidate results in no residual, it is immediately selected as a “skip”. Otherwise the merge candidates are tested as part of motion estimation when searching for the least cost inter option. The max candidate number is encoded in the SPS and determines the bit cost of signaling merge CUs. Default 2

--me <integer|string>
Motion search method. Generally, the higher the number the harder the ME method will try to find an optimal match. Diamond search is the simplest. Hexagon search is a little better. Uneven Multi-Hexagon is an adaption of the search method used by x264 for slower presets. Star is a three-step search adapted from the HM encoder: a star-pattern search followed by an optional radix scan followed by an optional star-search refinement. Full is an exhaustive search; an order of magnitude slower than all other searches but not much better than umh or star. SEA is similar to x264’s ESA implementation and a speed optimization of full search.

It is a three-step motion search where the DC calculation is followed by ADS calculation followed by SAD of the passed motion vector candidates.

  1. dia
  2. hex (default)
  3. umh
  4. star
  5. sea
  6. full
--subme, -m <0..7>

Amount of subpel refinement to perform. The higher the number the more subpel iterations and steps are performed. Default 2

-m HPEL iters HPEL dirs QPEL iters QPEL dirs HPEL SATD
0 1 4 0 4 false
1 1 4 1 4 false
2 1 4 1 4 true
3 2 4 1 4 true
4 2 4 2 4 true
5 1 8 1 8 true
6 2 8 1 8 true
7 2 8 2 8 true

At –subme values larger than 2, chroma residual cost is included in all subpel refinement steps and chroma residual is included in all motion estimation decisions (selecting the best reference picture in each list, and choosing between merge, uni-directional motion and bi-directional motion). The ‘slow’ preset is the first preset to enable the use of chroma residual.

--merange <integer>

Motion search range. Default 57

The default is derived from the default CTU size (64) minus the luma interpolation half-length (4) minus maximum subpel distance (2) minus one extra pixel just in case the hex search method is used. If the search range were any larger than this, another CTU row of latency would be required for reference frames.

Range of values: an integer from 0 to 32768

--temporal-mvp, --no-temporal-mvp

Enable temporal motion vector predictors in P and B slices. This enables the use of the motion vector from the collocated block in the previous frame to be used as a predictor. Default is enabled

--weightp, -w, --no-weightp

Enable weighted prediction in P slices. This enables weighting analysis in the lookahead, which influences slice decisions, and enables weighting analysis in the main encoder which allows P reference samples to have a weight function applied to them prior to using them for motion compensation. In video which has lighting changes, it can give a large improvement in compression efficiency. Default is enabled

--weightb, --no-weightb

Enable weighted prediction in B slices. Default disabled

--analyze-src-pics, --no-analyze-src-pics

Enable motion estimation with source frame pixels, in this mode, motion estimation can be computed independently. Default disabled.

--hme, --no-hme

Enable 3-level Hierarchical motion estimation at One-Sixteenth, Quarter and Full resolution. Default disabled.

Motion search method for HME Level 0, 1 and 2. Refer to --me for values. Specify search method for each level. Alternatively, specify a single value which will apply to all levels. Default is hex,umh,umh for levels 0,1,2 respectively.

--hme-range <integer>,<integer>,<integer>

Search range for HME level 0, 1 and 2. The Search Range for each HME level must be between 0 and 32768(excluding). Default search range is 16,32,48 for level 0,1,2 respectively.

Spatial/intra options

--strong-intra-smoothing, --no-strong-intra-smoothing

Enable strong intra smoothing for 32x32 intra blocks. This flag performs bi-linear interpolation of the corner reference samples for a strong smoothing effect. The purpose is to prevent blocking or banding artifacts in regions with few/zero AC coefficients. Default enabled

--constrained-intra, --no-constrained-intra

Constrained intra prediction. When generating intra predictions for blocks in inter slices, only intra-coded reference pixels are used. Inter-coded reference pixels are replaced with intra-coded neighbor pixels or default values. The general idea is to block the propagation of reference errors that may have resulted from lossy signals. Default disabled

Psycho-visual options

Left to its own devices, the encoder will make mode decisions based on a simple rate distortion formula, trading distortion for bitrate. This is generally effective except for the manner in which this distortion is measured. It tends to favor blurred reconstructed blocks over blocks which have wrong motion. The human eye generally prefers the wrong motion over the blur and thus x265 offers psycho-visual adjustments to the rate distortion algorithm.

--psy-rd will add an extra cost to reconstructed blocks which do not match the visual energy of the source block. The higher the strength of --psy-rd the more strongly it will favor similar energy over blur and the more aggressively it will ignore rate distortion. If it is too high, it will introduce visual artifacts and increase bitrate enough for rate control to increase quantization globally, reducing overall quality. psy-rd will tend to reduce the use of blurred prediction modes, like DC and planar intra and bi-directional inter prediction.

--psy-rdoq will adjust the distortion cost used in rate-distortion optimized quantization (RDO quant), enabled by --rdoq-level 1 or 2, favoring the preservation of energy in the reconstructed image. --psy-rdoq prevents RDOQ from blurring all of the encoding options which psy-rd has to choose from. At low strength levels, psy-rdoq will influence the quantization level decisions, favoring higher AC energy in the reconstructed image. As psy-rdoq strength is increased, more non-zero coefficient levels are added, and fewer coefficients are zeroed by RDOQ’s rate distortion analysis. High levels of psy-rdoq can double the bitrate which can have a drastic effect on rate control, forcing higher overall QP, and can cause ringing artifacts. psy-rdoq is less accurate than psy-rd, it is biasing towards energy in general while psy-rd biases towards the energy of the source image. But very large psy-rdoq values can sometimes be beneficial.

As a general rule, when both psycho-visual features are disabled, the encoder will tend to blur blocks in areas of difficult motion. Turning on small amounts of psy-rd and psy-rdoq will improve the perceived visual quality. Increasing psycho-visual strength further will improve quality and begin introducing artifacts and increase bitrate, which may force rate control to increase global QP. Finding the optimal psycho-visual parameters for a given video requires experimentation. Our recommended defaults (1.0 for both) are generally on the low end of the spectrum.

The lower the bitrate, the lower the optimal psycho-visual settings. If the bitrate is too low for the psycho-visual settings, you will begin to see temporal artifacts (motion judder). This is caused when the encoder is forced to code skip blocks (no residual) in areas of difficult motion because it is the best option psycho-visually (they have great amounts of energy and no residual cost). One can lower psy-rd settings when judder is happening and allow the encoder to use some blur in these areas of high motion.

In 444, chroma gets twice as much resolution, so halve the quality when psy-rd is enabled. So, when psy-rd is enabled for 444 videos, cbQpOffset and crQpOffset are set to value 6, if they are not explicitly set.

--psy-rd <float>

Influence rate distortion optimized mode decision to preserve the energy of the source image in the encoded image at the expense of compression efficiency. It only has effect on presets which use RDO-based mode decisions (--rd 3 and above). 1.0 is a typical value. Default 2.0

Range of values: 0 .. 5.0

--psy-rdoq <float>

Influence rate distortion optimized quantization by favoring higher energy in the reconstructed image. This generally improves perceived visual quality at the cost of lower quality metric scores. It only has effect when --rdoq-level is 1 or 2. High values can be beneficial in preserving high-frequency detail. Default: 0.0 (1.0 for presets slow, slower, veryslow)

Range of values: 0 .. 50.0

Slice decision options

--open-gop, --no-open-gop

Enable open GOP, allow I-slices to be non-IDR. Default enabled

--keyint, -I <integer>

Max intra period in frames. A special case of infinite-gop (single keyframe at the beginning of the stream) can be triggered with argument -1. Use 1 to force all-intra. When intra-refresh is enabled it specifies the interval between which refresh sweeps happen. Default 250

--min-keyint, -i <integer>

Minimum GOP size. Scenecuts beyond this interval are coded as IDR and start a new keyframe, while scenecuts closer together are coded as I or P. For fixed keyframe interval, set value to be equal to keyint.

Range of values: >=0 (0: auto)

--scenecut <integer>, --no-scenecut

How aggressively I-frames need to be inserted. The higher the threshold value, the more aggressive the I-frame placement. --scenecut 0 or --no-scenecut disables adaptive I frame placement. Default 40

--scenecut-bias <0..100.0>

This value represents the percentage difference between the inter cost and intra cost of a frame used in scenecut detection. For example, a value of 5 indicates, if the inter cost of a frame is greater than or equal to 95 percent of the intra cost of the frame, then detect this frame as scenecut. Values between 5 and 15 are recommended. Default 5.

--hist-scenecut, --no-hist-scenecut

Indicates that scenecuts need to be detected using luma edge and chroma histograms. --hist-scenecut enables scenecut detection using the histograms and disables the default scene cut algorithm. --no-hist-scenecut disables histogram based scenecut algorithm.

--hist-threshold <0.0..1.0>

This value represents the threshold for normalized SAD of edge histograms used in scenecut detection. This requires --hist-scenecut to be enabled. For example, a value of 0.2 indicates that a frame with normalized SAD value greater than 0.2 against the previous frame as scenecut. Increasing the threshold reduces the number of scenecuts detected. Default 0.03.

--radl <integer>

Number of RADL pictures allowed infront of IDR. Requires closed gop interval. If enabled for fixed keyframe interval, inserts RADL at every IDR. If enabled for closed gop interval, in case of --hist-scenecut inserts RADL at every hard scenecut whereas for the --scenecut, inserts RADL at every scenecut. Recommended value is 2-3. Default 0 (disabled).

**Range of values: Between 0 and –bframes

--ctu-info <0, 1, 2, 4, 6>

This value enables receiving CTU information asynchronously and determine reaction to the CTU information. Default 0. 1: force the partitions if CTU information is present. 2: functionality of (1) and reduce qp if CTU information has changed. 4: functionality of (1) and force Inter modes when CTU Information has changed, merge/skip otherwise. This option should be enabled only when planning to invoke the API function x265_encoder_ctu_info to copy ctu-info asynchronously. If enabled without calling the API function, the encoder will wait indefinitely.

--intra-refresh

Enables Periodic Intra Refresh(PIR) instead of keyframe insertion. PIR can replace keyframes by inserting a column of intra blocks in non-keyframes, that move across the video from one side to the other and thereby refresh the image but over a period of multiple frames instead of a single keyframe.

--rc-lookahead <integer>

Number of frames for slice-type decision lookahead (a key determining factor for encoder latency). The longer the lookahead buffer the more accurate scenecut decisions will be, and the more effective cutree will be at improving adaptive quant. Having a lookahead larger than the max keyframe interval is not helpful. Default 20

Range of values: Between the maximum consecutive bframe count (--bframes) and 250

--gop-lookahead <integer>

Number of frames for GOP boundary decision lookahead. If a scenecut frame is found within this from the gop boundary set by –keyint, the GOP will be extended until such a point, otherwise the GOP will be terminated as set by –keyint. Default 0.

Range of values: Between 0 and (–rc-lookahead - mini-GOP length)

It is recommended to have –gop-lookahaed less than –min-keyint as scenecuts beyond –min-keyint are already being coded as keyframes.

--lookahead-slices <0..16>

Use multiple worker threads to measure the estimated cost of each frame within the lookahead. The frame is divided into the specified number of slices, and one-thread is launched per slice. When --b-adapt is 2, most frame cost estimates will be performed in batch mode (many cost estimates at the same time) and lookahead-slices is ignored for batched estimates; it may still be used for single cost estimations. The higher this parameter, the less accurate the frame costs will be (since context is lost across slice boundaries) which will result in less accurate B-frame and scene-cut decisions. The effect on performance can be significant especially on systems with many threads.

The encoder may internally lower the number of slices or disable slicing to ensure each slice codes at least 10 16x16 rows of lowres blocks to minimize the impact on quality. For example, for 720p and 1080p videos, the number of slices is capped to 4 and 6, respectively. For resolutions lesser than 720p, slicing is auto-disabled.

If slices are used in lookahead, they are logged in the list of tools as lslices

Values: 0 - disabled. 1 is the same as 0. Max 16. Default: 8 for ultrafast, superfast, faster, fast, medium

4 for slow, slower disabled for veryslow, slower
--lookahead-threads <integer>
Use multiple worker threads dedicated to doing only lookahead instead of sharing the worker threads with frame Encoders. A dedicated lookahead threadpool is created with the specified number of worker threads. This can range from 0 upto half the hardware threads available for encoding. Using too many threads for lookahead can starve resources for frame Encoder and can harm performance. Default is 0 - disabled, Lookahead shares worker threads with other FrameEncoders .

Values: 0 - disabled(default). Max - Half of available hardware threads.

--b-adapt <integer>

Set the level of effort in determining B frame placement.

With b-adapt 0, the GOP structure is fixed based on the values of --keyint and --bframes.

With b-adapt 1 a light lookahead is used to choose B frame placement.

With b-adapt 2 (trellis) a viterbi B path selection is performed

Values: 0:none; 1:fast; 2:full(trellis) default

--bframes, -b <0..16>

Maximum number of consecutive b-frames. Use --bframes 0 to force all P/I low-latency encodes. Default 4. This parameter has a quadratic effect on the amount of memory allocated and the amount of work performed by the full trellis version of --b-adapt lookahead.

--bframe-bias <integer>

Bias towards B frames in slicetype decision. The higher the bias the more likely x265 is to use B frames. Can be any value between -90 and 100 and is clipped to that range. Default 0

--b-pyramid, --no-b-pyramid

Use B-frames as references, when possible. Default enabled

--force-flush <integer>

Force the encoder to flush frames. Default is 0.

Values: 0 - flush the encoder only when all the input pictures are over. 1 - flush all the frames even when the input is not over.

slicetype decision may change with this option.

2 - flush the slicetype decided frames only.

--fades, --no-fades

Detect and handle fade-in regions. Default disabled.

Quality, rate control and rate distortion options

--bitrate <integer>

Enables single-pass ABR rate control. Specify the target bitrate in kbps. Default is 0 (CRF)

Range of values: An integer greater than 0

--crf <0..51.0>

Quality-controlled variable bitrate. CRF is the default rate control method; it does not try to reach any particular bitrate target, instead it tries to achieve a given uniform quality and the size of the bitstream is determined by the complexity of the source video. The higher the rate factor the higher the quantization and the lower the quality. Default rate factor is 28.0.

--crf-max <0..51.0>

Specify an upper limit to the rate factor which may be assigned to any given frame (ensuring a max QP). This is dangerous when CRF is used in combination with VBV as it may result in buffer underruns. Default disabled

--crf-min <0..51.0>

Specify a lower limit to the rate factor which may be assigned to any given frame (ensuring a min compression factor).

--vbv-bufsize <integer>

Specify the size of the VBV buffer (kbits). Enables VBV in ABR mode. In CRF mode, --vbv-maxrate must also be specified. Default 0 (vbv disabled)

--vbv-maxrate <integer>

Maximum local bitrate (kbits/sec). Will be used only if vbv-bufsize is also non-zero. Both vbv-bufsize and vbv-maxrate are required to enable VBV in CRF mode. Default 0 (disabled)

Note that when VBV is enabled (with a valid --vbv-bufsize), VBV emergency denoising is turned on. This will turn on aggressive denoising at the frame level when frame QP > QP_MAX_SPEC (51), drastically reducing bitrate and allowing ratecontrol to assign lower QPs for the following frames. The visual effect is blurring, but removes significant blocking/displacement artifacts.

--vbv-init <float>

Initial buffer occupancy. The portion of the decode buffer which must be full before the decoder will begin decoding. Determines absolute maximum frame size. May be specified as a fractional value between 0 and 1, or in kbits. In other words, these two option pairs are equivalent:

--vbv-bufsize 1000 --vbv-init 900
--vbv-bufsize 1000 --vbv-init 0.9

Default 0.9

Range of values: fractional: 0 - 1.0, or kbits: 2 .. bufsize

--vbv-end <float>

Final buffer fullness. The portion of the decode buffer that must be full after all the specified frames have been inserted into the decode buffer. Specified as a fractional value between 0 and 1, or in kbits. Default 0 (disabled)

This enables basic support for chunk-parallel encoding where each segment can specify the starting and ending state of the VBV buffer so that VBV compliance can be maintained when chunks are independently encoded and stitched together.

--vbv-end-fr-adj <float>

Frame from which qp has to be adjusted to achieve final decode buffer fullness. Specified as a fraction of the total frames. Fractions > 0 are supported only when the total number of frames is known. Default 0.

--min-vbv-fullness <double>

Minimum VBV fullness percentage to be maintained. Specified as a fractional value ranging between 0 and 100. Default 50 i.e, Tries to keep the buffer at least 50% full at any point in time.

Decreasing the minimum required fullness shall improve the compression efficiency, but is expected to affect VBV conformance. Experimental option.
--max-vbv-fullness <double>

Maximum VBV fullness percentage to be maintained. Specified as a fractional value ranging between 0 and 100. Default 80 i.e Tries to keep the buffer at max 80% full at any point in time.

Increasing the minimum required fullness shall improve the compression efficiency,
but is expected to affect VBV conformance. Experimental option.
--qp, -q <integer>

Specify base quantization parameter for Constant QP rate control. Using this option enables Constant QP rate control. The specified QP is assigned to P slices. I and B slices are given QPs relative to P slices using param->rc.ipFactor and param->rc.pbFactor unless QP 0 is specified, in which case QP 0 is used for all slice types. Note that QP 0 does not cause lossless encoding, it only disables quantization. Default disabled.

Range of values: an integer from 0 to 51

--lossless, --no-lossless

Enables true lossless coding by bypassing scaling, transform, quantization and in-loop filter processes. This is used for ultra-high bitrates with zero loss of quality. Reconstructed output pictures are bit-exact to the input pictures. Lossless encodes implicitly have no rate control, all rate control options are ignored. Slower presets will generally achieve better compression efficiency (and generate smaller bitstreams). Default disabled.

--aq-mode <0|1|2|3|4>

Adaptive Quantization operating mode. Raise or lower per-block quantization based on complexity analysis of the source image. The more complex the block, the more quantization is used. These offsets the tendency of the encoder to spend too many bits on complex areas and not enough in flat areas.

  1. disabled
  2. AQ enabled
  3. AQ enabled with auto-variance (default)

3. AQ enabled with auto-variance and bias to dark scenes. This is recommended for 8-bit encodes or low-bitrate 10-bit encodes, to prevent color banding/blocking. 4. AQ enabled with auto-variance and edge information.

--aq-strength <float>

Adjust the strength of the adaptive quantization offsets. Setting --aq-strength to 0 disables AQ. At aq-modes 2 and 3, high aq-strengths will lead to high QP offsets resulting in a large difference in achieved bitrates.

Default 1.0. Range of values: 0.0 to 3.0

--hevc-aq

Enable adaptive quantization It scales the quantization step size according to the spatial activity of one coding unit relative to frame average spatial activity. This AQ method utilizes the minimum variance of sub-unit in each coding unit to represent the spatial complexity of the coding unit.

--qp-adaptation-range

Delta-QP range by QP adaptation based on a psycho-visual model.

Default 1.0. Range of values: 1.0 to 6.0

--aq-motion, --no-aq-motion

Adjust the AQ offsets based on the relative motion of each block with respect to the motion of the frame. The more the relative motion of the block, the more quantization is used. Default disabled. Experimental Feature

--qg-size <64|32|16|8>

Enable adaptive quantization for sub-CTUs. This parameter specifies the minimum CU size at which QP can be adjusted, ie. Quantization Group size. Allowed range of values are 64, 32, 16, 8 provided this falls within the inclusive range [maxCUSize, minCUSize]. Default: same as maxCUSize

--cutree, --no-cutree

Enable the use of lookahead’s lowres motion vector fields to determine the amount of reuse of each block to tune adaptive quantization factors. CU blocks which are heavily reused as motion reference for later frames are given a lower QP (more bits) while CU blocks which are quickly changed and are not referenced are given less bits. This tends to improve detail in the backgrounds of video with less detail in areas of high motion. Default enabled

--pass <integer>

Enable multi-pass rate control mode. Input is encoded multiple times, storing the encoded information of each pass in a stats file from which the consecutive pass tunes the qp of each frame to improve the quality of the output. Default disabled

  1. First pass, creates stats file
  2. Last pass, does not overwrite stats file
  3. Nth pass, overwrites stats file

Range of values: 1 to 3

--stats <filename>

Specify file name of of the multi-pass stats file. If unspecified the encoder will use x265_2pass.log

--slow-firstpass, --no-slow-firstpass

Enable first pass encode with the exact settings specified. The quality in subsequent multi-pass encodes is better (compared to first pass) when the settings match across each pass. Default enabled.

When slow first pass is disabled, a turbo encode with the following go-fast options is used to improve performance:

--multi-pass-opt-analysis, --no-multi-pass-opt-analysis

Enable/Disable multipass analysis refinement along with multipass ratecontrol. Based on the information stored in pass 1, in subsequent passes analysis data is refined and also redundant steps are skipped. In pass 1 analysis information like motion vector, depth, reference and prediction modes of the final best CTU partition is stored for each CTU. Multipass analysis refinement cannot be enabled when --analysis-save/analysis-load is enabled and both will be disabled when enabled together. This feature requires --pmode/--pme to be disabled and hence pmode/pme will be disabled when enabled at the same time.

Default: disabled.

--multi-pass-opt-distortion, --no-multi-pass-opt-distortion

Enable/Disable multipass refinement of qp based on distortion data along with multipass ratecontrol. In pass 1 distortion of best CTU partition is stored. CTUs with high distortion get lower(negative)qp offsets and vice-versa for low distortion CTUs in pass 2. This helps to improve the subjective quality. Multipass refinement of qp cannot be enabled when --analysis-save/--analysis-load is enabled and both will be disabled when enabled together. It requires --pmode/--pme to be disabled and hence pmode/pme will be disabled when enabled along with it.

Default: disabled.

--strict-cbr, --no-strict-cbr

Enables stricter conditions to control bitrate deviance from the target bitrate in ABR mode. Bit rate adherence is prioritised over quality. Rate tolerance is reduced to 50%. Default disabled.

This option is for use-cases which require the final average bitrate to be within very strict limits of the target; preventing overshoots, while keeping the bit rate within 5% of the target setting, especially in short segment encodes. Typically, the encoder stays conservative, waiting until there is enough feedback in terms of encoded frames to control QP. strict-cbr allows the encoder to be more aggressive in hitting the target bitrate even for short segment videos.

--cbqpoffs <integer>

Offset of Cb chroma QP from the luma QP selected by rate control. This is a general way to spend more or less bits on the chroma channel. Default 0

Range of values: -12 to 12

--crqpoffs <integer>

Offset of Cr chroma QP from the luma QP selected by rate control. This is a general way to spend more or less bits on the chroma channel. Default 0

Range of values: -12 to 12

--ipratio <float>

QP ratio factor between I and P slices. This ratio is used in all of the rate control modes. Some --tune options may change the default value. It is not typically manually specified. Default 1.4

--pbratio <float>

QP ratio factor between P and B slices. This ratio is used in all of the rate control modes. Some --tune options may change the default value. It is not typically manually specified. Default 1.3

--qcomp <float>

qComp sets the quantizer curve compression factor. It weights the frame quantizer based on the complexity of residual (measured by lookahead). It’s value must be between 0.5 and 1.0. Default value is 0.6. Increasing it to 1.0 will effectively generate CQP.

--qpstep <integer>

The maximum single adjustment in QP allowed to rate control. Default 4

--qpmin <integer>

sets a hard lower limit on QP allowed to ratecontrol. Default 0

--qpmax <integer>

sets a hard upper limit on QP allowed to ratecontrol. Default 69

--rc-grain, --no-rc-grain

Enables a specialised ratecontrol algorithm for film grain content. This parameter strictly minimises QP fluctuations within and across frames and removes pulsing of grain. Default disabled. Enabled when :option:’–tune’ grain is applied. It is highly recommended that this option is used through the tune grain feature where a combination of param options are used to improve visual quality.

--const-vbv, --no-const-vbv

Enables VBV algorithm to be consistent across runs. Default disabled. Enabled when :option:’–tune’ grain is applied.

--qblur <float>

Temporally blur quants. Default 0.5

--cplxblur <float>

temporally blur complexity. default 20

--zones <zone0>/<zone1>/...

Tweak the bitrate of regions of the video. Each zone takes the form:

<start frame>,<end frame>,<option> where <option> is either q=<integer> (force QP) or b=<float> (bitrate multiplier).

If zones overlap, whichever comes later in the list takes precedence. Default none

--zonefile <filename>

Specify a text file which contains the boundaries of the zones where each of zones are configurable. The format of each line is:

<frame number> <options to be configured>

The frame number indicates the beginning of a zone. The options following this is applied until another zone begins. The reconfigurable options can be specified as –<feature name> <feature value>

CLI ONLY

--scenecut-aware-qp, --no-scenecut-aware-qp

It reduces the bits spent on the inter-frames within the --scenecut-window before and after a scenecut by increasing their QP in ratecontrol pass2 algorithm without any deterioration in visual quality. If a scenecut falls within the window, the QP of the inter-frames after this scenecut will not be modified. --scenecut-aware-qp works only with –pass 2. Default disabled.

--scenecut-window <integer>

The duration(in milliseconds) for which there is a reduction in the bits spent on the inter-frames after a scenecut by increasing their QP, when --scenecut-aware-qp is enabled. Default 500ms.

Range of values: 0 to 1000

--qp-delta-ref <double>

The offset by which QP is incremented for inter-frames when --scenecut-aware-qp is enabled. Default 5.

Range of values: 0 to 10

--qp-delta-nonref <double>

The offset by which QP is incremented for non-referenced inter-frames when --scenecut-aware-qp is enabled. The offset is computed from --qp-delta-ref when it is not explicitly specified.

Range of values: 0 to 10

--vbv-live-multi-pass, --no-vbv-live-multi-pass

It enables the Qp tuning at frame level based on real time VBV Buffer fullness in the ratecontrol 2nd pass of multi pass mode to reduce the VBV violations. It could only be enabled with rate control stat-read encodes with VBV and ABR rate control mode.

Default disabled. Experimental feature

Quantization Options

Note that rate-distortion optimized quantization (RDOQ) is enabled implicitly at --rd 4, 5, and 6 and disabled implicitly at all other levels.

--signhide, --no-signhide

Hide sign bit of one coeff per TU (rdo). The last sign is implied. This requires analyzing all the coefficients to determine if a sign must be toggled, and then to determine which one can be toggled with the least amount of distortion. Default enabled

--qpfile <filename>

Specify a text file which contains frametypes and QPs for some or all frames. The format of each line is:

framenumber frametype QP

Frametype can be one of [I,i,K,P,B,b]. B is a referenced B frame, b is an unreferenced B frame. I is a keyframe (random access point) while i is an I frame that is not a keyframe (references are not broken). K implies I if closed_gop option is enabled, and i otherwise.

Specifying QP (integer) is optional, and if specified they are clamped within the encoder to qpmin/qpmax.

--scaling-list <filename>

Quantization scaling lists. HEVC supports 6 quantization scaling lists to be defined; one each for Y, Cb, Cr for intra prediction and one each for inter prediction.

x265 does not use scaling lists by default, but this can also be made explicit by --scaling-list off.

HEVC specifies a default set of scaling lists which may be enabled without requiring them to be signaled in the SPS. Those scaling lists can be enabled via --scaling-list default.

All other strings indicate a filename containing custom scaling lists in the HM format. The encode will abort if the file is not parsed correctly. Custom lists must be signaled in the SPS. A sample scaling list file is available in the downloads page

--lambda-file <filename>

Specify a text file containing values for x265_lambda_tab and x265_lambda2_tab. Each table requires MAX_MAX_QP+1 (70) float values.

The text file syntax is simple. Comma is considered to be white-space. All white-space is ignored. Lines must be less than 2k bytes in length. Content following hash (#) characters are ignored. The values read from the file are logged at --log-level debug.

Note that the lambda tables are process-global and so the new values affect all encoders running in the same process.

Lambda values affect encoder mode decisions, the lower the lambda the more bits it will try to spend on signaling information (motion vectors and splits) and less on residual. This feature is intended for experimentation.

--max-ausize-factor <float>

It controls the maximum AU size defined in specification. It represents the percentage of maximum AU size used. Default is 1. Range is 0.5 to 1.

Loop filters

--deblock=<int>:<int>, --no-deblock

Toggle deblocking loop filter, optionally specify deblocking strength offsets.

<int>:<int> - parsed as tC offset and Beta offset <int>,<int> - parsed as tC offset and Beta offset <int> - both tC and Beta offsets assigned the same value

If unspecified, the offsets default to 0. The offsets must be in a range of -6 (lowest strength) to 6 (highest strength).

To disable the deblocking filter entirely, use –no-deblock or –deblock=false. Default enabled, with both offsets defaulting to 0

If deblocking is disabled, or the offsets are non-zero, these changes from the default configuration are signaled in the PPS.

--sao, --no-sao

Toggle Sample Adaptive Offset loop filter, default enabled

--sao-non-deblock, --no-sao-non-deblock

Specify how to handle dependency between SAO and deblocking filter. When enabled, non-deblocked pixels are used for SAO analysis. When disabled, SAO analysis skips the right/bottom boundary areas. Default disabled

--limit-sao, --no-limit-sao

Limit SAO filter computation by early terminating SAO process based on inter prediction mode, CTU spatial-domain correlations, and relations between luma and chroma. Default disabled

--selective-sao <0..4>

Toggles SAO at slice level. Default 0.

Level Description
0 Disable SAO for all slices
1 Enable SAO only for I-slices
2 Enable SAO for I-slices & P-slices
3 Enable SAO for all reference slices
4 Enable SAO for all slices

VUI (Video Usability Information) options

x265 emits a VUI with only the timing info by default. If the SAR is specified (or read from a Y4M header) it is also included. All other VUI fields must be manually specified.

--sar <integer|w:h>

Sample Aspect Ratio, the ratio of width to height of an individual sample (pixel). The user may supply the width and height explicitly or specify an integer from the predefined list of aspect ratios defined in the HEVC specification. Default undefined (not signaled)

  1. 1:1 (square)
  2. 12:11
  3. 10:11
  4. 16:11
  5. 40:33
  6. 24:11
  7. 20:11
  8. 32:11
  9. 80:33
  10. 18:11
  11. 15:11
  12. 64:33
  13. 160:99
  14. 4:3
  15. 3:2
  16. 2:1
--display-window <left,top,right,bottom>

Define the (overscan) region of the image that does not contain information because it was added to achieve certain resolution or aspect ratio (the areas are typically black bars). The decoder may be directed to crop away this region before displaying the images via the --overscan option. Default undefined (not signaled).

Note that this has nothing to do with padding added internally by the encoder to ensure the pictures size is a multiple of the minimum coding unit (4x4). That padding is signaled in a separate “conformance window” and is not user-configurable.

--overscan <show|crop>

Specify whether it is appropriate for the decoder to display or crop the overscan area. Default unspecified (not signaled)

--videoformat <integer|string>

Specify the source format of the original analog video prior to digitizing and encoding. Default undefined (not signaled)

  1. component
  2. pal
  3. ntsc
  4. secam
  5. mac
  6. unknown
--range <full|limited>

Specify output range of black level and range of luma and chroma signals. Default undefined (not signaled)

--colorprim <integer|string>

Specify color primaries to use when converting to RGB. Default undefined (not signaled)

  1. bt709
  2. unknown
  3. reserved
  4. bt470m
  5. bt470bg
  6. smpte170m
  7. smpte240m
  8. film
  9. bt2020
  10. smpte428
  11. smpte431
  12. smpte432
--transfer <integer|string>

Specify transfer characteristics. Default undefined (not signaled)

  1. bt709
  2. unknown
  3. reserved
  4. bt470m
  5. bt470bg
  6. smpte170m
  7. smpte240m
  8. linear
  9. log100
  10. log316
  11. iec61966-2-4
  12. bt1361e
  13. iec61966-2-1
  14. bt2020-10
  15. bt2020-12
  16. smpte2084
  17. smpte428
  18. arib-std-b67
--colormatrix <integer|string>

Specify color matrix setting i.e set the matrix coefficients used in deriving the luma and chroma. Default undefined (not signaled)

  1. gbr
  2. bt709
  3. unknown
  4. reserved
  5. fcc
  6. bt470bg
  7. smpte170m
  8. smpte240m
  9. ycgco
  10. bt2020nc
  11. bt2020c
  12. smpte2085
  13. chroma-derived-nc
  14. chroma-derived-c
  15. ictcp
--chromaloc <0..5>

Specify chroma sample location for 4:2:0 inputs. Consult the HEVC specification for a description of these values. Default undefined (not signaled)

--master-display <string>

SMPTE ST 2086 mastering display color volume SEI info, specified as a string which is parsed when the stream header SEI are emitted. The string format is “G(%hu,%hu)B(%hu,%hu)R(%hu,%hu)WP(%hu,%hu)L(%u,%u)” where %hu are unsigned 16bit integers and %u are unsigned 32bit integers. The SEI includes X,Y display primaries for RGB channels and white point (WP) in units of 0.00002 and max,min luminance (L) values in units of 0.0001 candela per meter square. Applicable for HDR content.

Example for a P3D65 1000-nits monitor, where G(x=0.265, y=0.690), B(x=0.150, y=0.060), R(x=0.680, y=0.320), WP(x=0.3127, y=0.3290), L(max=1000, min=0.0001):

G(13250,34500)B(7500,3000)R(34000,16000)WP(15635,16450)L(10000000,1)

Note that this string value will need to be escaped or quoted to protect against shell expansion on many platforms. No default.

--max-cll <string>

Maximum content light level (MaxCLL) and maximum frame average light level (MaxFALL) as required by the Consumer Electronics Association 861.3 specification.

Specified as a string which is parsed when the stream header SEI are emitted. The string format is “%hu,%hu” where %hu are unsigned 16bit integers. The first value is the max content light level (or 0 if no maximum is indicated), the second value is the maximum picture average light level (or 0). Applicable for HDR content.

Example for MaxCLL=1000 candela per square meter, MaxFALL=400 candela per square meter:

–max-cll “1000,400”

Note that this string value will need to be escaped or quoted to protect against shell expansion on many platforms. No default.

--cll, --no-cll

Emit content light level SEI. Enabled automatically when --dolby-vision-profile 8.1 is specified. When enabled, signals max-cll and max-fall as 0 if max-cll is unspecified. Default enabled.

--hdr10, --no-hdr10

Force signaling of HDR10 parameters in SEI packets. Enabled automatically when --master-display or --max-cll is specified. Useful when there is a desire to signal 0 values for max-cll and max-fall. Default disabled.

--hdr10-opt, --no-hdr10-opt

Enable block-level luma and chroma QP optimization for HDR10 content as suggested in ITU-T H-series Recommendations ? Supplement 15. Source video should have HDR10 characteristics such as 10-bit depth 4:2:0 with Bt.2020 color primaries and SMPTE ST.2084 transfer characteristics. It is recommended that AQ-mode be enabled along with this feature. Default disabled.

--dhdr10-info <filename>

Inserts tone mapping information as an SEI message. It takes as input, the path to the JSON file containing the Creative Intent Metadata to be encoded as Dynamic Tone Mapping into the bitstream.

Click here for the syntax of the metadata file. A sample JSON file is available in the downloads page

--dhdr10-opt, --no-dhdr10-opt

Limits the frames for which tone mapping information is inserted as SEI message. Inserts SEI only for IDR frames and for frames where tone mapping information has changed.

--min-luma <integer>

Minimum luma value allowed for input pictures. Any values below min-luma are clipped. No default.

--max-luma <integer>

Maximum luma value allowed for input pictures. Any values above max-luma are clipped. No default.

--nalu-file <filename>

Text file containing userSEI in POC order : <POC><space><PREFIX><space><NAL UNIT TYPE>/<SEI TYPE><space><SEI Payload> Parse the input file specified and inserts SEI messages into the bitstream. Currently, we support only PREFIX SEI messages. This is an “application-only” feature.

--atc-sei <integer>

Emit the alternative transfer characteristics SEI message where the integer is the preferred transfer characteristics. Required for HLG (Hybrid Log Gamma) signaling. Not signaled by default.

--pic-struct <integer>

Set the picture structure and emits it in the picture timing SEI message. Values in the range 0..12. See D.3.3 of the HEVC spec. for a detailed explanation. Required for HLG (Hybrid Log Gamma) signaling. Not signaled by default.

Bitstream options

--annexb, --no-annexb

If enabled, x265 will produce Annex B bitstream format, which places start codes before NAL. If disabled, x265 will produce file format, which places length before NAL. x265 CLI will choose the right option based on output format. Default enabled

API ONLY

--repeat-headers, --no-repeat-headers

If enabled, x265 will emit VPS, SPS, and PPS headers with every keyframe. This is intended for use when you do not have a container to keep the stream headers for you and you want keyframes to be random access points. Default disabled

--aud, --no-aud

Emit an access unit delimiter NAL at the start of each slice access unit. If --repeat-headers is not enabled (indicating the user will be writing headers manually at the start of the stream) the very first AUD will be skipped since it cannot be placed at the start of the access unit, where it belongs. Default disabled

--hrd, --no-hrd

Enable the signaling of HRD parameters to the decoder. The HRD parameters are carried by the Buffering Period SEI messages and Picture Timing SEI messages providing timing information to the decoder. Default disabled

--hrd-concat, --no-hrd-concat

Set concatenation flag for the first keyframe in the HRD buffering period SEI. This is to signal the decoder if splicing is performed during bitstream generation. Recommended to enable this option during chunked encoding, except for the first chunk. Default disabled.

--dolby-vision-profile <integer|float>

Generate bitstreams confirming to the specified Dolby Vision profile, note that 0x7C01 makes RPU appear to be an unspecified NAL type in HEVC stream. If BL is backward compatible, Dolby Vision single layer VES will be equivalent to a backward compatible BL VES on legacy device as RPU will be ignored.

The value is specified as a float or as an integer with the profile times 10, for example profile 5 is specified as “5” or “5.0” or “50”.

Currently only profile 5, profile 8.1 and profile 8.2 enabled, Default 0 (disabled)

--dolby-vision-rpu <filename>

File containing Dolby Vision RPU metadata. If given, x265’s Dolby Vision metadata parser will fill the RPU field of input pictures with the metadata read from the file. The library will interleave access units with RPUs in the bitstream. Default NULL (disabled).

CLI ONLY

--info, --no-info

Emit an informational SEI with the stream headers which describes the encoder version, build info, and encode parameters. This is very helpful for debugging purposes but encoding version numbers and build info could make your bitstreams diverge and interfere with regression testing. Default enabled

--hash <integer>

Emit decoded picture hash SEI, so the decoder may validate the reconstructed pictures and detect data loss. Also useful as a debug feature to validate the encoder state. Default None

  1. MD5
  2. CRC
  3. Checksum
--temporal-layers,--no-temporal-layers

Enable a temporal sub layer. All referenced I/P/B frames are in the base layer and all unreferenced B frames are placed in a temporal enhancement layer. A decoder may choose to drop the enhancement layer and only decode and display the base layer slices.

If used with a fixed GOP (--b-adapt 0) and --bframes 3 then the two layers evenly split the frame rate, with a cadence of PbBbP. You probably also want --no-scenecut and a keyframe interval that is a multiple of 4.

--log2-max-poc-lsb <integer>

Maximum of the picture order count. Default 8

--vui-timing-info, --no-vui-timing-info

Emit VUI timing info in bitstream. Default enabled.

--vui-hrd-info, --no-vui-hrd-info

Emit VUI HRD info in bitstream. Default enabled when --hrd is enabled.

--opt-qp-pps, --no-opt-qp-pps

Optimize QP in PPS (instead of default value of 26) based on the QP values observed in last GOP. Default disabled.

--opt-ref-list-length-pps, --no-opt-ref-list-length-pps

Optimize L0 and L1 ref list length in PPS (instead of default value of 0) based on the lengths observed in the last GOP. Default disabled.

--multi-pass-opt-rps, --no-multi-pass-opt-rps

Enable storing commonly used RPS in SPS in multi pass mode. Default disabled.

--opt-cu-delta-qp, --no-opt-cu-delta-qp

Optimize CU level QPs by pulling up lower QPs to value close to meanQP thereby minimizing fluctuations in deltaQP signaling. Default disabled.

Only effective at RD levels 5 and 6

--idr-recovery-sei, --no-idr-recovery-sei

Emit RecoveryPoint info as sei in bitstream for each IDR frame. Default disabled.

--single-sei, --no-single-sei

Emit SEI messages in a single NAL unit instead of multiple NALs. Default disabled. When HRD SEI is enabled the HM decoder will throw a warning.

DCT Approximations

--lowpass-dct

If enabled, x265 will use low-pass subband dct approximation instead of the standard dct for 16x16 and 32x32 blocks. This approximation is less computationally intensive but it generates truncated coefficient matrixes for the transformed block. Empirical analysis shows marginal loss in compression and performance gains up to 10%, particularly at moderate bit-rates.

This approximation should be considered for platforms with performance and time constrains.

Default disabled. Experimental feature

Debugging options

--recon, -r <filename>

Output file containing reconstructed images in display order. If the file extension is “.y4m” the file will contain a YUV4MPEG2 stream header and frame headers. Otherwise it will be a raw YUV file in the encoder’s internal bit depth.

CLI ONLY

--recon-depth <integer>

Bit-depth of output file. This value defaults to the internal bit depth and currently cannot to be modified.

CLI ONLY

--recon-y4m-exec <string>

If you have an application which can play a Y4MPEG stream received on stdin, the x265 CLI can feed it reconstructed pictures in display order. The pictures will have no timing info, obviously, so the picture timing will be determined primarily by encoding elapsed time and latencies, but it can be useful to preview the pictures being output by the encoder to validate input settings and rate control parameters.

Example command for ffplay (assuming it is in your PATH):

–recon-y4m-exec “ffplay -i pipe:0 -autoexit”

CLI ONLY

ABR-ladder Options

--abr-ladder <filename>

File containing the encoder configurations to generate ABR ladder. The format of each line is:

<encID:reuse-level:refID> <CLI>

where, encID indicates the unique name given to the encode, refID indicates the name of the encode from which analysis info has to be re-used ( set to ‘nil’ if analysis reuse isn’t preferred ), and reuse-level indicates the level ( --analysis-load-reuse-level) at which analysis info has to be reused.

Sample config file:

[540p:0:nil] --input 540pSource.y4m --ctu 16 --bitrate 1600 --vbv-maxrate 2400 --vbv-bufsize 4800 -o 540p.hevc --preset veryslow
[1080p:10:540p] --input 1080pSource.y4m --ctu 32 --bitrate 5800 --vbv-maxrate 8700 --vbv-bufsize 17400 -o 1080p.hevc --preset veryslow --scale-factor 2
[2160p:10:1080p] --input 2160pSource.y4m --bitrate 16800 --vbv-maxrate 25200  --vbv-bufsize 50400 -o 2160p.hevc --preset veryslow  --scale-factor 2

The above sample config file is available in the downloads page

Default: Disabled ( Conventional single encode generation ). Experimental feature. CLI ONLY

SVT-HEVC Encoder Options

This section lists options which are SVT-HEVC encoder specific. See section svthevc for more details.

--svt, --no-svt

Enable SVT-HEVC encoder if x265 is built with SVT-HEVC library. Default disabled.

--svt-hme, --no-svt-hme

Enable Hierarchical Motion Estimation(HME) in SVT-HEVC. Default enabled.

CLI_ONLY

--svt-search-width <integer>

Search Area Width used during motion estimation. It depends on input resolution. Values: [1-256]

CLI_ONLY

--svt-search-height <integer>

Search Area Height used during motion estimation. It depends on input resolution. Values: [1-256]

CLI_ONLY

--svt-compressed-ten-bit-format, --no-svt-compressed-ten-bit-format

In order to reduce the size of input YUV and to increase channel density, SVT-HEVC accepts inputs in compressed-ten-bit-format. The conversion between yuv420p10le and compressed ten-bit format is a lossless operation. For more details about the conversion refer `here<https://github.com/intel/SVT-HEVC/blob/master/Docs/SVT-HEVC_Encoder_User_Guide.pdf>’_.

CLI_ONLY

--svt-speed-control, --no-svt-speed-control

Enable speed control functionality to achieve real time encoding speed defined by --fps. Default disabled.

CLI_ONLY

--svt-preset-tuner <integer>

SVT-HEVC exposes 12 presets. Presets [2-11] of SVT-HEVC is mapped to x265’s presets [placebo-ultrafast]. Ultrafast is mapped to preset(11) of SVT-HEVC, superfast to preset(10), placebo to preset(2) and so on. svt-preset-tuner works only on top of placebo preset and maps to presets (0-1) of SVT-HEVC.

Values: [0-1]

CLI_ONLY

--svt-hierarchical-level <integer>

Enables multiple hierarchical levels in SVT-HEVC. Accepts values in the range [0-3]. 0 - Flat 1 - 2-Level Hierarchy 2 - 3-Level Hierarchy 3 - 4-Level Hierarchy

Default: 3

CLI_ONLY

--svt-base-layer-switch-mode <integer>

Choose type of slices to be in base layer. Accepts values 0,1. 0 - Use B-frames in the base layer 1 - Use P-frames in the base layer

Default: 0

CLI_ONLY

--svt-pred-struct <integer>

Prediction structure forms the basis in deciding the GOP structure. SVT-HEVC supports Low delay(P/B) and random access prediction structure. In a low delay structure, pictures within a mini-gop can only refer to the previous pictures in display order. In other words, picture with display order N can only refer to pictures of display order lower than N. In random access method, pictures can be referenced from both the directions. It accepts values in the range [0-2]

0 - Low Delay P 1 - Low Delay B 2 - Random Access

Default: 2

CLI_ONLY

--svt-fps-in-vps, --no-svt-fps-in-vps

Enable sending timing info in VPS. Default disabled.

CLI_ONLY

Application Programming Interface

Introduction

x265 is written primarily in C++ and x86 assembly language but the public facing programming interface is C for the widest possible portability. This C interface is wholly defined within x265.h in the source/ folder of our source tree. All of the functions and variables and enumerations meant to be used by the end-user are present in this header.

Where possible, x265 has tried to keep its public API as close as possible to x264’s public API. So those familiar with using x264 through its C interface will find x265 quite familiar.

This file is meant to be read in-order; the narrative follows linearly through the various sections

Build Considerations

The choice of Main or Main10 profile encodes is made at compile time; the internal pixel depth influences a great deal of variable sizes and thus 8 and 10bit pixels are handled as different build options (primarily to maintain the performance of the 8bit builds). libx265 exports a variable x265_max_bit_depth which indicates how the library was compiled (it will contain a value of 8 or 10). Further, x265_version_str is a pointer to a string indicating the version of x265 which was compiled, and x265_build_info_str is a pointer to a string identifying the compiler and build options.

Note

x265_version_str is only updated when cmake runs. If you are making binaries for others to use, it is recommended to run cmake prior to make in your build scripts.

x265 will accept input pixels of any depth between 8 and 16 bits regardless of the depth of its internal pixels (8 or 10). It will shift and mask input pixels as required to reach the internal depth. If downshifting is being performed using our CLI application (to 8 bits), the --dither option may be enabled to reduce banding. This feature is not available through the C interface.

Encoder

The primary object in x265 is the encoder object, and this is represented in the public API as an opaque typedef x265_encoder. Pointers of this type are passed to most encoder functions.

A single encoder generates a single output bitstream from a sequence of raw input pictures. Thus if you need multiple output bitstreams you must allocate multiple encoders. You may pass the same input pictures to multiple encoders, the encode function does not modify the input picture structures (the pictures are copied into the encoder as the first step of encode).

Encoder allocation is a reentrant function, so multiple encoders may be safely allocated in a single process. The encoder access functions are not reentrant for a single encoder, so the recommended use case is to allocate one client thread per encoder instance (one thread for all encoder instances is possible, but some encoder access functions are blocking and thus this would be less efficient).

Note

There is one caveat to having multiple encoders within a single process. All of the encoders must use the same maximum CTU size because many global variables are configured based on this size. Encoder allocation will fail if a mis-matched CTU size is attempted. If no encoders are open, x265_cleanup() can be called to reset the configured CTU size so a new size can be used.

An encoder is allocated by calling x265_encoder_open():

/* x265_encoder_open:
 *      create a new encoder handler, all parameters from x265_param are copied */
x265_encoder* x265_encoder_open(x265_param *);

The returned pointer is then passed to all of the functions pertaining to this encode. A large amount of memory is allocated during this function call, but the encoder will continue to allocate memory as the first pictures are passed to the encoder; until its pool of picture structures is large enough to handle all of the pictures it must keep internally. The pool size is determined by the lookahead depth, the number of frame threads, and the maximum number of references.

As indicated in the comment, x265_param is copied internally so the user may release their copy after allocating the encoder. Changes made to their copy of the param structure have no affect on the encoder after it has been allocated.

Param

The x265_param structure describes everything the encoder needs to know about the input pictures and the output bitstream and most everything in between.

The recommended way to handle these param structures is to allocate them from libx265 via:

/* x265_param_alloc:
 *  Allocates an x265_param instance. The returned param structure is not
 *  special in any way, but using this method together with x265_param_free()
 *  and x265_param_parse() to set values by name allows the application to treat
 *  x265_param as an opaque data struct for version safety */
x265_param *x265_param_alloc();

In this way, your application does not need to know the exact size of the param structure (the build of x265 could potentially be a bit newer than the copy of x265.h that your application compiled against).

Next you perform the initial rough cut configuration of the encoder by chosing a performance preset and optional tune factor x265_preset_names and x265_tune_names respectively hold the string names of the presets and tune factors (see presets for more detail on presets and tune factors):

/*      returns 0 on success, negative on failure (e.g. invalid preset/tune name). */
int x265_param_default_preset(x265_param *, const char *preset, const char *tune);

Now you may optionally specify a profile. x265_profile_names contains the string names this function accepts:

/*      (can be NULL, in which case the function will do nothing)
 *      returns 0 on success, negative on failure (e.g. invalid profile name). */
int x265_param_apply_profile(x265_param *, const char *profile);

Finally you configure any remaining options by name using repeated calls to:

/* x265_param_parse:
 *  set one parameter by name.
 *  returns 0 on success, or returns one of the following errors.
 *  note: BAD_VALUE occurs only if it can't even parse the value,
 *  numerical range is not checked until x265_encoder_open().
 *  value=NULL means "true" for boolean options, but is a BAD_VALUE for non-booleans. */
#define X265_PARAM_BAD_NAME  (-1)
#define X265_PARAM_BAD_VALUE (-2)
int x265_param_parse(x265_param *p, const char *name, const char *value);

See string options for the list of options (and their descriptions) which can be set by x265_param_parse().

After the encoder has been created, you may release the param structure:

/* x265_param_free:
 *  Use x265_param_free() to release storage for an x265_param instance
 *  allocated by x265_param_alloc() */
void x265_param_free(x265_param *);

Note

Using these methods to allocate and release the param structures helps future-proof your code in many ways, but the x265 API is versioned in such a way that we prevent linkage against a build of x265 that does not match the version of the header you are compiling against (unless you use x265_api_query() to acquire the library’s interfaces). This is function of the X265_BUILD macro.

x265_encoder_parameters() may be used to get a copy of the param structure from the encoder after it has been opened, in order to see the changes made to the parameters for auto-detection and other reasons:

/* x265_encoder_parameters:
 *      copies the current internal set of parameters to the pointer provided
 *      by the caller.  useful when the calling application needs to know
 *      how x265_encoder_open has changed the parameters.
 *      note that the data accessible through pointers in the returned param struct
 *      (e.g. filenames) should not be modified by the calling application. */
void x265_encoder_parameters(x265_encoder *, x265_param *);

x265_encoder_reconfig() may be used to reconfigure encoder parameters mid-encode:

    /* x265_encoder_reconfig:
     *       used to modify encoder parameters.
     *      various parameters from x265_param are copied.
     *      this takes effect immediately, on whichever frame is encoded next;
     *      returns negative on parameter validation error, 0 on successful reconfigure
     *      and 1 when a reconfigure is already in progress.
     *
     *      not all parameters can be changed; see the actual function for a
     *      detailed breakdown.  since not all parameters can be changed, moving
     *      from preset to preset may not always fully copy all relevant parameters,
     *      but should still work usably in practice. however, more so than for
     *      other presets, many of the speed shortcuts used in ultrafast cannot be
     *      switched out of; using reconfig to switch between ultrafast and other
     *      presets is not recommended without a more fine-grained breakdown of
     *      parameters to take this into account. */
int x265_encoder_reconfig(x265_encoder *, x265_param *);

x265_encoder_reconfig_zone() Used to reconfigure rate-contol settings of zones mid-encode:

/* x265_encoder_reconfig_zone:
 *      Properties of the zone will be copied to encoder's param and will be used only to
 *      influence rate-control decisions of the zone.
 *      returns 0 on successful copy and negative on failure.*/
int x265_encoder_reconfig(x265_encoder *, x265_param *);

x265_get_slicetype_poc_and_scenecut() may be used to fetch slice type, poc and scene cut information mid-encode:

/* x265_get_slicetype_poc_and_scenecut:
 *     get the slice type, poc and scene cut information for the current frame,
 *     returns negative on error, 0 on success.
 *     This API must be called after(poc >= lookaheadDepth + bframes + 2) condition check. */
 int x265_get_slicetype_poc_and_scenecut(x265_encoder *encoder, int *slicetype, int *poc, int* sceneCut);

x265_get_ref_frame_list() may be used to fetch forward and backward refrence list:

/* x265_get_ref_frame_list:
 *     returns negative on error, 0 when access unit were output.
 *     This API must be called after(poc >= lookaheadDepth + bframes + 2) condition check */
 int x265_get_ref_frame_list(x265_encoder *encoder, x265_picyuv**, x265_picyuv**, int, int, int*, int*);

x265_encoder_ctu_info may be used to provide additional CTU-specific information to the encoder:

/* x265_encoder_ctu_info:
 *    Copy CTU information such as ctu address and ctu partition structure of all
 *    CTUs in each frame. The function is invoked only if "--ctu-info" is enabled and
 *    the encoder will wait for this copy to complete if enabled.*/
int x265_encoder_ctu_info(x265_encoder *encoder, int poc, x265_ctu_info_t** ctu);

x265_set_analysis_data() may be used to recive analysis information from external application:

/* x265_set_analysis_data:
 *     set the analysis data. The incoming analysis_data structure is assumed to be AVC-sized blocks.
 *     returns negative on error, 0 access unit were output.*/
 int x265_set_analysis_data(x265_encoder *encoder, x265_analysis_data *analysis_data, int poc, uint32_t cuBytes);

x265_alloc_analysis_data() may be used to allocate memory for the x265_analysis_data:

/* x265_alloc_analysis_data:
 *     Allocate memory for the x265_analysis_data object's internal structures. */
 void x265_alloc_analysis_data(x265_param *param, x265_analysis_data* analysis);

x265_free_analysis_data() may be used to free memory for the x265_analysis_data:

/* x265_free_analysis_data:
 *    Free the allocated memory for x265_analysis_data object's internal structures. */
 void x265_free_analysis_data(x265_param *param, x265_analysis_data* analysis);

Pictures

Raw pictures are passed to the encoder via the x265_picture structure. Just like the param structure we recommend you allocate this structure from the encoder to avoid potential size mismatches:

/* x265_picture_alloc:
 *  Allocates an x265_picture instance. The returned picture structure is not
 *  special in any way, but using this method together with x265_picture_free()
 *  and x265_picture_init() allows some version safety. New picture fields will
 *  always be added to the end of x265_picture */
x265_picture *x265_picture_alloc();

Regardless of whether you allocate your picture structure this way or whether you simply declare it on the stack, your next step is to initialize the structure via:

/***
 * Initialize an x265_picture structure to default values. It sets the pixel
 * depth and color space to the encoder's internal values and sets the slice
 * type to auto - so the lookahead will determine slice type.
 */
void x265_picture_init(x265_param *param, x265_picture *pic);

x265 does not perform any color space conversions, so the raw picture’s color space (chroma sampling) must match the color space specified in the param structure used to allocate the encoder. x265_picture_init initializes this field to the internal color space and it is best to leave it unmodified.

The picture bit depth is initialized to be the encoder’s internal bit depth but this value should be changed to the actual depth of the pixels being passed into the encoder. If the picture bit depth is more than 8, the encoder assumes two bytes are used to represent each sample (little-endian shorts).

The user is responsible for setting the plane pointers and plane strides (in units of bytes, not pixels). The presentation time stamp (pts) is optional, depending on whether you need accurate decode time stamps (dts) on output.

If you wish to override the lookahead or rate control for a given picture you may specify a slicetype other than X265_TYPE_AUTO, or a forceQP value other than 0.

x265 does not modify the picture structure provided as input, so you may reuse a single x265_picture for all pictures passed to a single encoder, or even all pictures passed to multiple encoders.

Structures allocated from the library should eventually be released:

/* x265_picture_free:
 *  Use x265_picture_free() to release storage for an x265_picture instance
 *  allocated by x265_picture_alloc() */
void x265_picture_free(x265_picture *);

Analysis Buffers

Analysis information can be saved and reused to between encodes of the same video sequence (generally for multiple bitrate encodes). The best results are attained by saving the analysis information of the highest bitrate encode and reuse it in lower bitrate encodes.

When saving or loading analysis data, buffers must be allocated for every picture passed into the encoder using:

/* x265_alloc_analysis_data:
 *  Allocate memory to hold analysis meta data, returns 1 on success else 0 */
int x265_alloc_analysis_data(x265_picture*);

Note that this is very different from the typical semantics of x265_picture, which can be reused many times. The analysis buffers must be re-allocated for every input picture.

Analysis buffers passed to the encoder are owned by the encoder until they pass the buffers back via an output x265_picture. The user is responsible for releasing the buffers when they are finished with them via:

/* x265_free_analysis_data:
 *  Use x265_free_analysis_data to release storage of members allocated by
 *  x265_alloc_analysis_data */
void x265_free_analysis_data(x265_picture*);

Encode Process

The output of the encoder is a series of NAL packets, which are always returned concatenated in consecutive memory. HEVC streams have SPS and PPS and VPS headers which describe how the following packets are to be decoded. If you specified --repeat-headers then those headers will be output with every keyframe. Otherwise you must explicitly query those headers using:

/* x265_encoder_headers:
 *      return the SPS and PPS that will be used for the whole stream.
 *      *pi_nal is the number of NAL units outputted in pp_nal.
 *      returns negative on error, total byte size of payload data on success
 *      the payloads of all output NALs are guaranteed to be sequential in memory. */
int x265_encoder_headers(x265_encoder *, x265_nal **pp_nal, uint32_t *pi_nal);

Now we get to the main encode loop. Raw input pictures are passed to the encoder in display order via:

/* x265_encoder_encode:
 *      encode one picture.
 *      *pi_nal is the number of NAL units outputted in pp_nal.
 *      returns negative on error, zero if no NAL units returned.
 *      the payloads of all output NALs are guaranteed to be sequential in memory. */
int x265_encoder_encode(x265_encoder *encoder, x265_nal **pp_nal, uint32_t *pi_nal, x265_picture *pic_in, x265_picture *pic_out);

These pictures are queued up until the lookahead is full, and then the frame encoders in turn are filled, and then finally you begin receiving a output NALs (corresponding to a single output picture) with each input picture you pass into the encoder.

Once the pipeline is completely full, x265_encoder_encode() will block until the next output picture is complete.

Note

Optionally, if the pointer of a second x265_picture structure is provided, the encoder will fill it with data pertaining to the output picture corresponding to the output NALs, including the recontructed image, POC and decode timestamp. These pictures will be in encode (or decode) order. The encoder will also write corresponding frame encode statistics into x265_frame_stats.

When the last of the raw input pictures has been sent to the encoder, x265_encoder_encode() must still be called repeatedly with a pic_in argument of 0, indicating a pipeline flush, until the function returns a value less than or equal to 0 (indicating the output bitstream is complete).

At any time during this process, the application may query running statistics from the encoder:

/* x265_encoder_get_stats:
 *       returns encoder statistics */
void x265_encoder_get_stats(x265_encoder *encoder, x265_stats *, uint32_t statsSizeBytes);

Cleanup

At the end of the encode, the application will want to trigger logging of the final encode statistics, if --csv had been specified:

/* x265_encoder_log:
 *       write a line to the configured CSV file. If a CSV filename was not
 *       configured, or file open failed, this function will perform no write. */
void x265_encoder_log(x265_encoder *encoder, int argc, char **argv);

Finally, the encoder must be closed in order to free all of its resources. An encoder that has been flushed cannot be restarted and reused. Once x265_encoder_close() has been called, the encoder handle must be discarded:

/* x265_encoder_close:
 *      close an encoder handler */
void x265_encoder_close(x265_encoder *);

When the application has completed all encodes, it should call x265_cleanup() to free process global, particularly if a memory-leak detection tool is being used. x265_cleanup() also resets the saved CTU size so it will be possible to create a new encoder with a different CTU size:

/* x265_cleanup:
 *     release library static allocations, reset configured CTU size */
void x265_cleanup(void);

VMAF (Video Multi-Method Assessment Fusion)

If you set the ENABLE_LIBVMAF cmake option to ON, then x265 will report per frame and aggregate VMAF score for the given input and dump the scores in csv file. The user also need to specify the --recon in command line to get the VMAF scores.

/* x265_calculate_vmafScore:
  • returns VMAF score for the input video.
  • This api must be called only after encoding was done. */

double x265_calculate_vmafscore(x265_param*, x265_vmaf_data*);

/* x265_calculate_vmaf_framelevelscore:
  • returns VMAF score for each frame in a given input video. The frame level VMAF score does not include temporal scores. */

double x265_calculate_vmaf_framelevelscore(x265_vmaf_framedata*);

Note

When setting ENABLE_LIBVMAF cmake option to ON, it is recommended to also set ENABLE_SHARED to OFF to prevent build problems. We only need the static library from these builds.

Binaries build with windows will not have VMAF support.

Multi-library Interface

If your application might want to make a runtime bit-depth selection, it will need to use one of these bit-depth introspection interfaces which returns an API structure containing the public function entry points and constants.

Instead of directly using all of the x265_ methods documented above, you query an x265_api structure from your libx265 and then use the function pointers of the same name (minus the x265_ prefix) within that structure. For instance x265_param_default() becomes api->param_default().

x265_api_get

The first bit-depth instrospecton method is x265_api_get(). It designed for applications that might statically link with libx265, or will at least be tied to a particular SONAME or API version:

/* x265_api_get:
 *   Retrieve the programming interface for a linked x265 library.
 *   May return NULL if no library is available that supports the
 *   requested bit depth. If bitDepth is 0, the function is guarunteed
 *   to return a non-NULL x265_api pointer from the system default
 *   libx265 */
const x265_api* x265_api_get(int bitDepth);

Like x265_encoder_encode(), this function has the build number automatically appended to the function name via macros. This ties your application to a particular binary API version of libx265 (the one you compile against). If you attempt to link with a libx265 with a different API version number, the link will fail.

Obviously this has no meaningful effect on applications which statically link to libx265.

x265_api_query

The second bit-depth introspection method is designed for applications which need more flexibility in API versioning. If you use x265_api_query() and dynamically link to libx265 at runtime (using dlopen() on POSIX or LoadLibrary() on Windows) your application is no longer directly tied to the API version that it was compiled against:

/* x265_api_query:
 *   Retrieve the programming interface for a linked x265 library, like
 *   x265_api_get(), except this function accepts X265_BUILD as the second
 *   argument rather than using the build number as part of the function name.
 *   Applications which dynamically link to libx265 can use this interface to
 *   query the library API and achieve a relative amount of version skew
 *   flexibility. The function may return NULL if the library determines that
 *   the apiVersion that your application was compiled against is not compatible
 *   with the library you have linked with.
 *
 *   api_major_version will be incremented any time non-backward compatible
 *   changes are made to any public structures or functions. If
 *   api_major_version does not match X265_MAJOR_VERSION from the x265.h your
 *   application compiled against, your application must not use the returned
 *   x265_api pointer.
 *
 *   Users of this API *must* also validate the sizes of any structures which
 *   are not treated as opaque in application code. For instance, if your
 *   application dereferences a x265_param pointer, then it must check that
 *   api->sizeof_param matches the sizeof(x265_param) that your application
 *   compiled with. */
const x265_api* x265_api_query(int bitDepth, int apiVersion, int* err);

A number of validations must be performed on the returned API structure in order to determine if it is safe for use by your application. If you do not perform these checks, your application is liable to crash:

if (api->api_major_version != X265_MAJOR_VERSION) /* do not use */
if (api->sizeof_param != sizeof(x265_param))      /* do not use */
if (api->sizeof_picture != sizeof(x265_picture))  /* do not use */
if (api->sizeof_stats != sizeof(x265_stats))      /* do not use */
if (api->sizeof_zone != sizeof(x265_zone))        /* do not use */
etc.

Note that if your application does not directly allocate or dereference one of these structures, if it treats the structure as opaque or does not use it at all, then it can skip the size check for that structure.

In particular, if your application uses api->param_alloc(), api->param_free(), api->param_parse(), etc and never directly accesses any x265_param fields, then it can skip the check on the sizeof(x265_parm) and thereby ignore changes to that structure (which account for a large percentage of X265_BUILD bumps).

Build Implications

By default libx265 will place all of its internal C++ classes and functions within an x265 namespace and export all of the C functions documented in this file. Obviously this prevents 8bit and 10bit builds of libx265 from being statically linked into a single binary, all of those symbols would collide.

However, if you set the EXPORT_C_API cmake option to OFF then libx265 will use a bit-depth specific namespace and prefix for its assembly functions (x265_8bit, x265_10bit or x265_12bit) and export no C functions.

In this way you can build one or more libx265 libraries without any exported C interface and link them into a libx265 build that does export a C interface. The build which exported the C functions becomes the default bit depth for the combined library, and the other bit depths are available via the bit-depth introspection methods.

Note

When setting EXPORT_C_API cmake option to OFF, it is recommended to also set ENABLE_SHARED and ENABLE_CLI to OFF to prevent build problems. We only need the static library from these builds.

If an application requests a bit-depth that is not supported by the default library or any of the additionally linked libraries, the introspection method will fall-back to an attempt to dynamically bind a shared library with a name appropriate for the requested bit-depth:

8-bit:  libx265_main
10-bit: libx265_main10
12-bit: libx265_main12

If the profile-named library is not found, it will then try to bind a generic libx265 in the hopes that it is a multilib library with all bit depths.

Packaging and Distribution

We recommend that packagers distribute a single combined shared/static library build which includes all the bit depth libraries linked together. See the multilib scripts in our build/ subdirectories for examples of how to affect these combined library builds. It is the packager’s discretion which bit-depth exports the public C functions and thus becomes the default bit-depth for the combined library.

Note

Windows packagers might want to build libx265 with WINXP_SUPPORT enabled. This makes the resulting binaries functional on XP and Vista. Without this flag, the minimum supported host O/S is Windows 7. Also note that binaries built with WINXP_SUPPORT will not have NUMA support and they will have slightly less performance.

STATIC_LINK_CRT is also recommended so end-users will not need to install any additional MSVC C runtime libraries.

Threading

Thread Pools

x265 creates one or more thread pools per encoder, one pool per NUMA node (typically a CPU socket). --pools specifies the number of pools and the number of threads per pool the encoder will allocate. By default x265 allocates one thread per (hyperthreaded) CPU core on each NUMA node.

If you are running multiple encoders on a system with multiple NUMA nodes, it is recommended to isolate each of them to a single node in order to avoid the NUMA overhead of remote memory access.

Work distribution is job based. Idle worker threads scan the job providers assigned to their thread pool for jobs to perform. When no jobs are available, the idle worker threads block and consume no CPU cycles.

Objects which desire to distribute work to worker threads are known as job providers (and they derive from the JobProvider class). The thread pool has a method to poke awake a blocked idle thread, and job providers are recommended to call this method when they make new jobs available.

Worker jobs are not allowed to block except when absolutely necessary for data locking. If a job becomes blocked, the work function is expected to drop that job so the worker thread may go back to the pool and find more work.

On Windows, the native APIs offer sufficient functionality to discover the NUMA topology and enforce the thread affinity that libx265 needs (so long as you have not chosen to target XP or Vista), but on POSIX systems it relies on libnuma for this functionality. If your target POSIX system is single socket, then building without libnuma is a perfectly reasonable option, as it will have no effect on the runtime behavior. On a multiple-socket system, a POSIX build of libx265 without libnuma will be less work efficient, but will still function correctly. You lose the work isolation effect that keeps each frame encoder from only using the threads of a single socket and so you incur a heavier context switching cost.

Wavefront Parallel Processing

New with HEVC, Wavefront Parallel Processing allows each row of CTUs to be encoded in parallel, so long as each row stays at least two CTUs behind the row above it, to ensure the intra references and other data of the blocks above and above-right are available. WPP has almost no effect on the analysis and compression of each CTU and so it has a very small impact on compression efficiency relative to slices or tiles. The compression loss from WPP has been found to be less than 1% in most of our tests.

WPP has three effects which can impact efficiency. The first is the row starts must be signaled in the slice header, the second is each row must be padded to an even byte in length, and the third is the state of the entropy coder is transferred from the second CTU of each row to the first CTU of the row below it. In some conditions this transfer of state actually improves compression since the above-right state may have better locality than the end of the previous row.

Parabola Research have published an excellent HEVC animation which visualizes WPP very well. It even correctly visualizes some of WPPs key drawbacks, such as:

  1. the low thread utilization at the start and end of each frame
  2. a difficult block may stall the wave-front and it takes a while for the wave-front to recover.
  3. 64x64 CTUs are big! there are much fewer rows than with H.264 and similar codecs

Because of these stall issues you rarely get the full parallelisation benefit one would expect from row threading. 30% to 50% of the theoretical perfect threading is typical.

In x265 WPP is enabled by default since it not only improves performance at encode but it also makes it possible for the decoder to be threaded.

If WPP is disabled by --no-wpp the frame will be encoded in scan order and the entropy overheads will be avoided. If frame threading is not disabled, the encoder will change the default frame thread count to be higher than if WPP was enabled. The exact formulas are described in the next section.

Bonded Task Groups

If a worker thread job has work which can be performed in parallel by many threads, it may allocate a bonded task group and enlist the help of other idle worker threads from the same thread pool. Those threads will cooperate to complete the work of the bonded task group and then return to their idle states. The larger and more uniform those tasks are, the better the bonded task group will perform.

Parallel Mode Analysis

When --pmode is enabled, each CU (at all depths from 64x64 to 8x8) will distribute its analysis work to the thread pool via a bonded task group. Each analysis job will measure the cost of one prediction for the CU: merge, skip, intra, inter (2Nx2N, Nx2N, 2NxN, and AMP).

At slower presets, the amount of increased parallelism from pmode is often enough to be able to reduce or disable frame parallelism while achieving the same overall CPU utilization. Reducing frame threads is often beneficial to ABR and VBV rate control.

Parallel Motion Estimation

When --pme is enabled all of the analysis functions which perform motion searches to reference frames will distribute those motion searches to other worker threads via a bonded task group (if more than two motion searches are required).

Frame Threading

Frame threading is the act of encoding multiple frames at the same time. It is a challenge because each frame will generally use one or more of the previously encoded frames as motion references and those frames may still be in the process of being encoded themselves.

Previous encoders such as x264 worked around this problem by limiting the motion search region within these reference frames to just one macroblock row below the coincident row being encoded. Thus a frame could be encoded at the same time as its reference frames so long as it stayed one row behind the encode progress of its references (glossing over a few details).

x265 has the same frame threading mechanism, but we generally have much less frame parallelism to exploit than x264 because of the size of our CTU rows. For instance, with 1080p video x264 has 68 16x16 macroblock rows available each frame while x265 only has 17 64x64 CTU rows.

The second extenuating circumstance is the loop filters. The pixels used for motion reference must be processed by the loop filters and the loop filters cannot run until a full row has been encoded, and it must run a full row behind the encode process so that the pixels below the row being filtered are available. On top of this, HEVC has two loop filters: deblocking and SAO, which must be run in series with a row lag between them. When you add up all the row lags each frame ends up being 3 CTU rows behind its reference frames (the equivalent of 12 macroblock rows for x264). And keep in mind the wave-front progression pattern; by the time the reference frame finishes the third row of CTUs, nearly half of the CTUs in the frame may be compressed (depending on the display aspect ratio).

The third extenuating circumstance is that when a frame being encoded becomes blocked by a reference frame row being available, that frame’s wave-front becomes completely stalled and when the row becomes available again it can take quite some time for the wave to be restarted, if it ever does. This makes WPP less effective when frame parallelism is in use.

--merange can have a negative impact on frame parallelism. If the range is too large, more rows of CTU lag must be added to ensure those pixels are available in the reference frames.

Note

Even though the merange is used to determine the amount of reference pixels that must be available in the reference frames, the actual motion search is not necessarily centered around the coincident block. The motion search is actually centered around the motion predictor, but the available pixel area (mvmin, mvmax) is determined by merange and the interpolation filter half-heights.

When frame threading is disabled, the entirety of all reference frames are always fully available (by definition) and thus the available pixel area is not restricted at all, and this can sometimes improve compression efficiency. Because of this, the output of encodes with frame parallelism disabled will not match the output of encodes with frame parallelism enabled; but when enabled the number of frame threads should have no effect on the output bitstream except when using ABR or VBV rate control or noise reduction.

When --nr is enabled, the outputs of each number of frame threads will be deterministic but none of them will match becaue each frame encoder maintains a cumulative noise reduction state.

VBV introduces non-determinism in the encoder, at this point in time, regardless of the amount of frame parallelism.

By default frame parallelism and WPP are enabled together. The number of frame threads used is auto-detected from the (hyperthreaded) CPU core count, but may be manually specified via --frame-threads

Cores Frames
> 32 6..8
>= 16 5
>= 8 3
>= 4 2

If WPP is disabled, then the frame thread count defaults to min(cpuCount, ctuRows / 2)

Over-allocating frame threads can be very counter-productive. They each allocate a large amount of memory and because of the limited number of CTU rows and the reference lag, you generally get limited benefit from adding frame encoders beyond the auto-detected count, and often the extra frame encoders reduce performance.

Given these considerations, you can understand why the faster presets lower the max CTU size to 32x32 (making twice as many CTU rows available for WPP and for finer grained frame parallelism) and reduce --merange

Each frame encoder runs in its own thread (allocated separately from the worker pool). This frame thread has some pre-processing responsibilities and some post-processing responsibilities for each frame, but it spends the bulk of its time managing the wave-front processing by making CTU rows available to the worker threads when their dependencies are resolved. The frame encoder threads spend nearly all of their time blocked in one of 4 possible locations:

  1. blocked, waiting for a frame to process
  2. blocked on a reference frame, waiting for a CTU row of reconstructed and loop-filtered reference pixels to become available
  3. blocked waiting for wave-front completion
  4. blocked waiting for the main thread to consume an encoded frame

Lookahead

The lookahead module of x265 (the lowres pre-encode which determines scene cuts and slice types) uses the thread pool to distribute the lowres cost analysis to worker threads. It will use bonded task groups to perform batches of frame cost estimates, and it may optionally use bonded task groups to measure single frame cost estimates using slices. (see --lookahead-slices)

The main slicetypeDecide() function itself is also performed by a worker thread if your encoder has a thread pool, else it runs within the context of the thread which calls the x265_encoder_encode().

SAO

The Sample Adaptive Offset loopfilter has a large effect on encode performance because of the peculiar way it must be analyzed and coded.

SAO flags and data are encoded at the CTU level before the CTU itself is coded, but SAO analysis (deciding whether to enable SAO and with what parameters) cannot be performed until that CTU is completely analyzed (reconstructed pixels are available) as well as the CTUs to the right and below. So in effect the encoder must perform SAO analysis in a wavefront at least a full row behind the CTU compression wavefront.

This extra latency forces the encoder to save the encode data of every CTU until the entire frame has been analyzed, at which point a function can code the final slice bitstream with the decided SAO flags and data interleaved between each CTU. This second pass over the CTUs can be expensive, particularly at large resolutions and high bitrates.

Preset Options

Presets

x265 has ten predefined --preset options that optimize the trade-off between encoding speed (encoded frames per second) and compression efficiency (quality per bit in the bitstream). The default preset is medium. It does a reasonably good job of finding the best possible quality without spending excessive CPU cycles looking for the absolute most efficient way to achieve that quality. When you use faster presets, the encoder takes shortcuts to improve performance at the expense of quality and compression efficiency. When you use slower presets, x265 tests more encoding options, using more computations to achieve the best quality at your selected bit rate (or in the case of –crf rate control, the lowest bit rate at the selected quality).

The presets adjust encoder parameters as shown in the following table. Any parameters below that are specified in your command-line will be changed from the value specified by the preset.

  1. ultrafast
  2. superfast
  3. veryfast
  4. faster
  5. fast
  6. medium (default)
  7. slow
  8. slower
  9. veryslow
  10. placebo
preset 0 1 2 3 4 5 6 7 8 9
ctu 32 32 64 64 64 64 64 64 64 64
min-cu-size 16 8 8 8 8 8 8 8 8 8
bframes 3 3 4 4 4 4 4 8 8 8
b-adapt 0 0 0 0 0 2 2 2 2 2
rc-lookahead 5 10 15 15 15 20 25 40 40 60
lookahead-slices 8 8 8 8 8 8 4 1 1 1
scenecut 0 40 40 40 40 40 40 40 40 40
ref 1 1 2 2 3 3 4 5 5 5
limit-refs 0 0 3 3 3 3 3 1 0 0
me dia hex hex hex hex hex star star star star
merange 57 57 57 57 57 57 57 57 57 92
subme 0 1 1 2 2 2 3 4 4 5
rect 0 0 0 0 0 0 1 1 1 1
amp 0 0 0 0 0 0 0 1 1 1
limit-modes 0 0 0 0 0 0 1 1 0 0
max-merge 2 2 2 2 2 2 3 4 5 5
early-skip 1 1 1 1 0 1 0 0 0 0
recursion-skip 1 1 1 1 1 1 1 1 1 0
fast-intra 1 1 1 1 1 0 0 0 0 0
b-intra 0 0 0 0 0 0 0 1 1 1
sao 0 0 1 1 1 1 1 1 1 1
signhide 0 1 1 1 1 1 1 1 1 1
weightp 0 0 1 1 1 1 1 1 1 1
weightb 0 0 0 0 0 0 0 1 1 1
aq-mode 0 0 2 2 2 2 2 2 2 2
cuTree 1 1 1 1 1 1 1 1 1 1
rdLevel 2 2 2 2 2 3 4 6 6 6
rdoq-level 0 0 0 0 0 0 2 2 2 2
tu-intra 1 1 1 1 1 1 1 3 3 4
tu-inter 1 1 1 1 1 1 1 3 3 4
limit-tu 0 0 0 0 0 0 0 4 0 0

Tuning

There are a few --tune options available, which are applied after the preset.

Note

The psnr and ssim tune options disable all optimizations that sacrafice metric scores for perceived visual quality (also known as psycho-visual optimizations). By default x265 always tunes for highest perceived visual quality but if one intends to measure an encode using PSNR or SSIM for the purpose of benchmarking, we highly recommend you configure x265 to tune for that particular metric.

–tune effect
psnr disables adaptive quant, psy-rd, and cutree
ssim enables adaptive quant auto-mode, disables psy-rd
grain improves retention of film grain. more below
fastdecode no loop filters, no weighted pred, no intra in B
zerolatency no lookahead, no B frames, no cutree
animation improves encode quality for animated content

Film Grain

--tune grain aims to encode grainy content with the best visual quality. The purpose of this option is neither to retain nor eliminate grain, but prevent noticeable artifacts caused by uneven distribution of grain. --tune grain strongly restricts algorithms that vary the quantization parameter within and across frames. Tune grain also biases towards decisions that retain more high frequency components.

It also enables a specialised ratecontrol algorithm --rc-grain that strictly minimises QP fluctuations across frames, while still allowing the encoder to hit bitrate targets and VBV buffer limits (with a slightly higher margin of error than normal). It is highly recommended that this algorithm is used only through the --tune grain feature. Overriding the –tune grain settings might result in grain strobing, especially when enabling features like --aq-mode and --cutree that modify per-block QPs within a given frame.

Fast Decode

--tune fastdecode disables encoder features which tend to be bottlenecks for the decoder. It is intended for use with 4K content at high bitrates which can cause decoders to struggle. It disables both HEVC loop filters, which tend to be process bottlenecks:

It disables weighted prediction, which tend to be bandwidth bottlenecks:

And it disables intra blocks in B frames with --no-b-intra since intra predicted blocks cause serial dependencies in the decoder.

Zero Latency

There are two halves to the latency problem. There is latency at the decoder and latency at the encoder. --tune zerolatency removes latency from both sides. The decoder latency is removed by:

Encoder latency is removed by:

With all of these settings x265_encoder_encode() will run synchronously, the picture passed as pic_in will be encoded and returned as NALs. These settings disable frame parallelism, which is an important component for x265 performance. If you can tolerate any latency on the encoder, you can increase performance by increasing the number of frame threads. Each additional frame thread adds one frame of latency.

Animation

--tune animation adjusts encoder settings to optimize the encode quality for animation content without impacting the encode speed. This is done by:

Lossless

Lossless Encoding

x265 can encode HEVC bitstreams that are entirely lossless (the reconstructed images are bit-exact to the source images) by using the --lossless option. Lossless operation is theoretically simple. Rate control, by definition, is disabled and the encoder disables all quality metrics since they would only waste CPU cycles. Instead, x265 reports only a compression factor at the end of the encode.

In HEVC, lossless coding means bypassing both the DCT transforms and bypassing quantization (often referred to as transquant bypass). Normal predictions are still allowed, so the encoder will find optimal inter or intra predictions and then losslessly code the residual (with transquant bypass).

All --preset options are capable of generating lossless video streams, but in general the slower the preset the better the compression ratio (and the slower the encode). Here are some examples:

./x265 ../test-720p.y4m o.bin --preset ultrafast --lossless
... <snip> ...
encoded 721 frames in 238.38s (3.02 fps), 57457.94 kb/s

./x265 ../test-720p.y4m o.bin --preset faster --lossless
... <snip> ...
x265 [info]: lossless compression ratio 3.11::1
encoded 721 frames in 258.46s (2.79 fps), 56787.65 kb/s

./x265 ../test-720p.y4m o.bin --preset slow --lossless
... <snip> ...
x265 [info]: lossless compression ratio 3.36::1
encoded 721 frames in 576.73s (1.25 fps), 52668.25 kb/s

./x265 ../test-720p.y4m o.bin --preset veryslow --lossless
x265 [info]: lossless compression ratio 3.76::1
encoded 721 frames in 6298.22s (0.11 fps), 47008.65 kb/s

Note

In HEVC, only QP=4 is truly lossless quantization, and thus when encoding losslesly x265 uses QP=4 internally in its RDO decisions.

Near-lossless Encoding

Near-lossless conditions are a quite a bit more interesting. Normal ABR rate control will allow one to scale the bitrate up to the point where quantization is entirely bypassed (QP <= 4), but even at this point there is a lot of SSIM left on the table because of the DCT transforms, which are not lossless:

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 40000 --ssim
encoded 721 frames in 326.62s (2.21 fps), 39750.56 kb/s, SSIM Mean Y: 0.9990703 (30.317 dB)

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 50000 --ssim
encoded 721 frames in 349.27s (2.06 fps), 44326.84 kb/s, SSIM Mean Y: 0.9994134 (32.316 dB)

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 60000 --ssim
encoded 721 frames in 360.04s (2.00 fps), 45394.50 kb/s, SSIM Mean Y: 0.9994823 (32.859 dB)

For the encoder to get over this quality plateau, one must enable lossless coding at the CU level with --cu-lossless. It tells the encoder to evaluate trans-quant bypass as a coding option for each CU, and to pick the option with the best rate-distortion characteristics.

The --cu-lossless option is very expensive, computationally, and it only has a positive effect when the QP is extremely low, allowing RDO to spend a large amount of bits to make small improvements to quality. So this option should only be enabled when you are encoding near-lossless bitstreams:

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 40000 --ssim --cu-lossless
encoded 721 frames in 500.51s (1.44 fps), 40017.10 kb/s, SSIM Mean Y: 0.9997790 (36.557 dB)

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 50000 --ssim --cu-lossless
encoded 721 frames in 524.60s (1.37 fps), 46083.37 kb/s, SSIM Mean Y: 0.9999432 (42.456 dB)

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 60000 --ssim --cu-lossless
encoded 721 frames in 523.63s (1.38 fps), 46552.92 kb/s, SSIM Mean Y: 0.9999489 (42.917 dB)

Note

It is not unusual for bitrate to drop as you increase lossless coding. Having “perfectly coded” reference blocks reduces residual in later frames. It is quite possible for a near-lossless encode to spend more bits than a lossless encode.

Enabling psycho-visual rate distortion will improve lossless coding. --psy-rd influences the RDO decisions in favor of energy (detail) preservation over bit cost and results in more blocks being losslessly coded. Our psy-rd feature is not yet assembly optimized, so this makes the encodes run even slower:

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 40000 --ssim --cu-lossless --psy-rd 1.0
encoded 721 frames in 581.83s (1.24 fps), 40112.15 kb/s, SSIM Mean Y: 0.9998632 (38.638 dB)

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 50000 --ssim --cu-lossless --psy-rd 1.0
encoded 721 frames in 587.54s (1.23 fps), 46284.55 kb/s, SSIM Mean Y: 0.9999663 (44.721 dB)

./x265 ../test-720p.y4m o.bin --preset medium --bitrate 60000 --ssim --cu-lossless --psy-rd 1.0
encoded 721 frames in 592.93s (1.22 fps), 46839.51 kb/s, SSIM Mean Y: 0.9999707 (45.334 dB)

--cu-lossless will also be more effective at slower presets which perform RDO at more levels and thus may find smaller blocks that would benefit from lossless coding:

./x265 ../test-720p.y4m o.bin --preset veryslow --bitrate 40000 --ssim --cu-lossless
encoded 721 frames in 12969.25s (0.06 fps), 37331.96 kb/s, SSIM Mean Y: 0.9998108 (37.231 dB)

./x265 ../test-720p.y4m o.bin --preset veryslow --bitrate 50000 --ssim --cu-lossless
encoded 721 frames in 46217.84s (0.05 fps), 42976.28 kb/s, SSIM Mean Y: 0.9999482 (42.856 dB)

./x265 ../test-720p.y4m o.bin --preset veryslow --bitrate 60000 --ssim --cu-lossless
encoded 721 frames in 13738.17s (0.05 fps), 43864.21 kb/s, SSIM Mean Y: 0.9999633 (44.348 dB)

And with psy-rd and a slow preset together, very high SSIMs are possible:

./x265 ../test-720p.y4m o.bin --preset veryslow --bitrate 40000 --ssim --cu-lossless --psy-rd 1.0
encoded 721 frames in 11675.81s (0.06 fps), 37819.45 kb/s, SSIM Mean Y: 0.9999181 (40.867 dB)

./x265 ../test-720p.y4m o.bin --preset veryslow --bitrate 50000 --ssim --cu-lossless --psy-rd 1.0
encoded 721 frames in 12414.56s (0.06 fps), 42815.75 kb/s, SSIM Mean Y: 0.9999758 (46.168 dB)

./x265 ../test-720p.y4m o.bin --preset veryslow --bitrate 60000 --ssim --cu-lossless --psy-rd 1.0
encoded 721 frames in 11684.89s (0.06 fps), 43324.48 kb/s, SSIM Mean Y: 0.9999793 (46.835 dB)

It’s important to note in the end that it is easier (less work) for the encoder to encode the video losslessly than it is to encode it near-losslessly. If the encoder knows up front the encode must be lossless, it does not need to evaluate any lossy coding methods. The encoder only needs to find the most efficient prediction for each block and then entropy code the residual.

It is not feasible for --cu-lossless to turn itself on when the encoder determines it is encoding a near-lossless bitstream (ie: when rate control nearly disables all quantization) because the feature requires a flag to be enabled in the stream headers. At the time the stream headers are being coded we do not know whether --cu-lossless would be a help or a hinder. If very few or no blocks end up being coded as lossless, then having the feature enabled is a net loss in compression efficiency because it adds a flag that must be coded for every CU. So ignoring even the performance aspects of the feature, it can be a compression loss if enabled without being used. So it is up to the user to only enable this feature when they are coding at near-lossless quality.

Transform Skip

A somewhat related feature, --tskip tells the encoder to evaluate transform-skip (bypass DCT but with quantization still enabled) when coding small 4x4 transform blocks. This feature is intended to improve the coding efficiency of screen content (aka: text on a screen) and is not really intended for lossless coding. This feature should only be enabled if the content has a lot of very sharp edges in it, and is mostly unrelated to lossless coding.

SVT-HEVC

x265 has support for open source HEVC encoder SVT-HEVC and can generate SVT-HEVC compliant bitstreams. SVT-HEVC encoder can be enabled at run time using --svt. Since SVT-HEVC params/CLI are not exposed outside, it has to be configured only via x265 CLI options. The API’s of SVT-HEVC are accessed through x265’s API so even library users of x265 can avail this feature, under the condition that x265_param_parse() API should be used for all param assignment. Params assigned outside of x265_param_parse() API wont’t be mapped to SVT-HEVC. This document describes the steps needed to compile x265 with SVT-HEVC and CLI options mapping between x265 and SVT-HEVC.

Supported Version

Version - 1.4.1

Build Steps

This section describes the build steps to be followed to link SVT-HEVC with x265.

SVT-HEVC

  1. Clone SVT-HEVC (say at path “/home/app/”) and build it (follow the build steps in its README file)
  2. Once build is successful, binaries can be found inside the Bin folder at its root directory (“/home/app/SVT-HEVC/Bin/Release/”)

x265

  1. Set environmental variables SVT_HEVC_INCLUDE_DIR and SVT_HEVC_LIBRARY_DIR to help x265 locate SVT-HEVC. For example:
    • export SVT_HEVC_INCLUDE_DIR = /home/app/SVT-HEVC/Source/API/
    • export SVT_HEVC_LIBRARY_DIR = /home/app/SVT-HEVC/Bin/Release/
  2. Enable the cmake option ENABLE_SVT_HEVC and continue the normal build process

CLI options Mapping

Once x265 is compiled with SVT-HEVC, SVT-HEVC encoder can be invoked at run time using --svt. --svt can be added anywhere in the command line, x265 application will automatically parse it first and maps all other x265 CLI’s present in the command line to SVT-HEVC encoder options internally and configures the same. Below table shows the actual mapping of x265 CLI options to SVT-HEVC encoder options (Name as shown in SVT-HEVC’s sample configuration file)

x265 CLI option SVT-HEVC Encoder Option Range
--input InputFile Any String
--output StreamFile Any String
--input-depth EncoderBitDepth [8, 10]
--input-res SourceWidth [64 - 8192]
SourceHeight [64 - 8192]
--frames FrameToBeEncoded Any Positive Integer
--fps FrameRateNumerator Any Positive Integer
FrameRateDenominator Any Positive Integer
--profile Profile [main, main10]
--level-idc Level
[1, 2, 2.1, 3, 3.1, 4, 4.1,
5, 5.1, 5.2, 6, 6.1, 6.2]
--high-tier Tier  
--qpmin MinQpAllowed [0 - 50]
--qpmax MaxQpAllowed [0 - 51]
--rc-lookahead LookAheadDistance [0 - 250]
--scenecut SceneChangeDetection Any Positive Integer
--open-gop IntraRefreshType  
--deblock LoopFilterDisable Any Integer
--sao SAO  
--keyint IntraPeriod [(-2) - 255]
--constrained-intra ConstrainedIntra  
--vui-timing-info VideoUsabilityInfo  
--hdr HighDynamicRangeInput  
--aud AccessUnitDelimeter  
--qp RateControlMode RateControlMode = 0
QP [0 - 51]
--bitrate RateControlMode RateControlMode = 1
TargetBitrate Any Positive Integer
--interlace InterlacedVideo [0 - 2]
--pools TargetSocket, LogicalProcessors Maximum NUMA Nodes = 2
--master-display MasteringDisplayColorVolume Any String
--max-cll maxCLL, maxFALL Any Positve Integer
--dolby-vision-profile DolbyVisionProfile [8.1]
--dolby-vision-rpu DolbyVisionRpuFile Any String
--nalu-file NaluFile Any String
--hrd hrdFlag [0, 1]
--vbv-maxrate vbvMaxrate Any Positive Integer
--vbv-bufsize vbvBufsize Any Positive Integer
--vbv-init VbvBufInit [0 - 100]
--frame-threads ThreadCount Any Number
--svt-search-width SearchAreaWidth [1 - 256]
--svt-search-height SearchAreaHeight [1 - 256]
--svt-hierarchical-level HierarchicalLevels [0 - 3]
--svt-base-layer-switch-mode BaseLayerSwitchMode [0, 1]
--svt-pred-struct PredStructure [0 - 2]
--svt-hme HME, UseDefaultMeHme  
--svt-compressed-ten-bit-format CompressedTenBitFormat  
--svt-speed-control SpeedControlFlag  
--svt-fps-in-vps FpsInVps  

x265 CLI options which are not present in the above table will have no effect if SVT-HEVC is enabled and would be ignored silently with a warning. If SVT-HEVC is enabled, accepted input range of x265 CLI options will change, because it follows SVT-HEVC encoder’s specs, which are mentioned in the Range section in the above table. Options starting with prefix “–svt-” are newly added to fecilitate access to the features of SVT-HEVC which couldn’t be mapped to the existing x265 CLI’s. So these options will have effect only if SVT-HEVC is enabled and would be ignored with default x265 encode.

Preset Option Mapping

x265 has 10 presets from ultrafast to placebo whereas SVT-HEVC has 12 presets. Use --svt-preset-tuner with Placebo preset to access the additional 2 presets of SVT-HEVC. Note that --svt-preset-tuner should be used only if SVT-HEVC is enabled and only with Placebo preset, would be ignored otherwise. Below table shows the actual mapping of presets,

x265 Preset SVT-HEVC Preset
Ultrafast 11
Superfast 10
Veryfast 9
Faster 8
Fast 7
Medium 6
Slow 5
Slower 4
Veryslow 3
Placebo 2
Placebo --svt-preset-tuner 0 0
Placebo --svt-preset-tuner 1 1

Release Notes

Version 3.4

Release date - 29th May, 2020.

New features

  1. Edge-aware quadtree partitioning to terminate CU depth recursion based on edge information. --rskip level 2 enables the feature and --rskip-edge-threshold denotes the minimum expected edge-density percentage within the CU, below which the recursion is skipped. Experimental feature.
  2. Application-level feature --abr-ladder for automating efficient ABR ladder generation. Shows ~65% savings in the over-all turn-around time required for the generation of a typical Apple HLS ladder in Intel(R) Xeon(R) Platinum 8280 CPU @ 2.70GHz over a sequential ABR-ladder generation approach that leverages save-load architecture.

Enhancements to existing features

  1. Improved efficiency in 2-pass rate-control algorithm. The savings in the bitrate is ~1.72% with visual improvement in quality in the initial 1-2 secs.

Encoder enhancements

  1. Faster ARM64 encodes enabled by ASM contributions from Huawei. The speed-up over no-asm version for 1080p encodes @ medium preset is ~15% in a 16 core H/W.
  2. Strict VBV conformance in zone encoding.

Bug fixes

  1. Multi-pass encode failures with --frame-dup.
  2. Corrupted bitstreams with --hist-scenecut when input depth and internal bit-depth differ.
  3. Incorrect analysis propagation in multi-level save-load architecture.
  4. Failure in detecting NUMA packages installed in non-standard directories.

Version 3.3

Release date - 17th February, 2020.

New features

  1. Adaptive frame duplication to identify and skip encoding of near-identical frames and signal the duplication info to the decoder via pic_struct SEI. frame-dup to enable frame duplication and --dup-threshold to set the threshold for frame similarity (optional).
  2. Boundary aware quantization to cut off bits from frames following scene-cut. This leverages the inability of HVS to perceive fine details during scene changes and saves bits. --scenecut-aware-qp , --scenecut-window and --max-qp-delta to enable boundary aware frame quantization, to set window size (optional) and to set QP offset (optional).
  3. Improved scene-cut detection using edge and chroma histograms. --hist-scenecut to enable the feature and --hist-threshold (optional) to provide threshold for determining scene cuts.

Enhancements to existing features

  1. --hme-range to modify search range for HME levels L0, L1, and L2.
  2. Improved performance of AQ mode 4 by reducing memory foot print.
  3. Introduced --analysis-save-reuse-level and --analysis-load-reuse-level to de-couple reuse levels of --analysis-save and --analysis-load. Turnaround time of ABR encoding can be reduced by properly leveraging these options.

Encoder enhancements

  1. Improved VBV lookahead to eliminate blocky artifacts in Intra frames coming towards end of the title.

API changes

  1. New API function x265_encoder_reconfig_zone() to invoke zone reconfiguration dynamically.
  2. Renamed --hdr to --hdr10. --hdr will be deprecated in the upcoming major release.
  3. Renamed --hdr-opt to --hdr10-opt. --hdr-opt will be deprecated in the upcoming major release.
  4. Additions to x265_param structure to support the newly added features and encoder enhancements.

Bug fixes

  1. Output change in --analysis-load at inter-refine levels 2 and 3.
  2. Encoder crash with zones.
  3. Integration issues with SVT v1.4.1.
  4. Fixed bug in --limit-tu 3 and 4 while loading co-located CU’s TU depth.

Version 3.2

Release date - 25th September, 2019.

New features

  1. 3-level hierarchical motion estimation using --hme and --hme-search.
  2. New AQ mode (--aq-mode 4) with variance and edge information.
  3. selective-sao to selectively enable SAO at slice level.

Enhancements to existing features

  1. New implementation of --refine-mv with 3 refinement levels.

Encoder enhancements

  1. Improved quality in the frames following dark scenes in ABR mode.

API changes

  1. Additions to x265_param structure to support the newly added features --hme, --hme-search and selective-sao.

Bug fixes

  1. Fixed encoder crash with --zonefile during failures in encoder_open().
  2. Fixed JSON11 build errors with HDR10+ on MacOS high sierra.
  3. Signalling out of range scaling list data fixed.
  4. Inconsistent output fix for 2-pass rate-control with cutree ON.

Known issues

  1. Build dependency on changeset cf37911 of SVT-HEVC.

Version 3.1

Release date - 18th June, 2019.

New features

  1. x265 can invoke SVT-HEVC library for encoding through --svt.
  2. x265 can now accept interlaced inputs directly (no need to separate fields), and sends it to the encoder with proper fps and frame-size through --field.
  3. --fades can detect and handle fade-in regions. This option will force I-slice and initialize RC history for the brightest frame after fade-in.

API changes

  1. A new flag to signal MasterDisplayParams and maxCll/Fall separately

Encoder enhancements

  1. Improved the performance of inter-refine level 1 by skipping the evaluation of smaller CUs when the current block is decided as “skip” by the save mode.
  2. New AVX2 primitives to improve the performance of encodes that enable --ssim-rd.
  3. Improved performance in medium preset with negligible loss in quality.

Bug fixes

  1. Bug fixes for zones.
  2. Fixed wrap-around from MV structure overflow occurred around 8K pixels or over.
  3. Fixed issues in configuring cbQpOffset and crQpOffset for 444 input
  4. Fixed cutree offset computation in 2nd pass encodes.

Known issues

  1. AVX512 main12 asm disabling.
  2. Inconsistent output with 2-pass due to cutree offset sharing.

Version 3.0

Release date - 23/01/2019

New features

  1. option:: ‘–dolby-vision-profile <integer|float>’ generates bitstreams confirming to the specified Dolby Vision profile. Currently profile 5, profile 8.1 and profile 8.2 enabled, Default 0 (disabled)
  2. option:: ‘–dolby-vision-rpu’ File containing Dolby Vision RPU metadata. If given, x265’s Dolby Vision metadata parser will fill the RPU field of input pictures with the metadata
    read from the file. The library will interleave access units with RPUs in the bitstream. Default NULL (disabled).
  3. option:: ‘–zonefile <filename>’ specifies a text file which contains the boundaries of the zones where each of zones are configurable.
  4. option:: ‘–qp-adaptation-range’ Delta-QP range by QP adaptation based on a psycho-visual model. Default 1.0.
  5. option:: ‘–refine-ctu-distortion <0/1>’ store/normalize ctu distortion in analysis-save/load. Default 0.
  6. Experimental feature option:: ‘–hevc-aq’ enables adaptive quantization
    It scales the quantization step size according to the spatial activity of one coding unit relative to frame average spatial activity. This AQ method utilizes the minimum variance of sub-unit in each coding unit to represent the coding unit’s spatial complexity.

Encoder enhancements

  1. Preset: change param defaults for veryslow and slower preset. Replace slower preset with defaults used in veryslow preset and change param defaults in veryslow preset as per experimental results.
  2. AQ: change default AQ mode to auto-variance
  3. Cutree offset reuse: restricted to analysis reuse-level 10 for analysis-save -> analysis-load
  4. Tune: introduce –tune animation option which improves encode quality for animated content
  5. Reuse CU depth for B frame and allow I, P frame to follow x265 depth decision

Bug fixes

  1. RC: fix rowStat computation in const-vbv
  2. Dynamic-refine: fix memory reset size.
  3. Fix Issue #442: linking issue on non x86 platform
  4. Encoder: Do not include CLL SEI message if empty
  5. Fix issue #441 build error in VMAF lib

Version 2.9

Release date - 05/10/2018

New features

  1. Support for chunked encoding

    --chunk-start and --chunk-end Frames preceding first frame of chunk in display order will be encoded, however, they will be discarded in the bitstream. Frames following last frame of the chunk in display order will be used in taking lookahead decisions, but, they will not be encoded. This feature can be enabled only in closed GOP structures. Default disabled.

  2. Support for HDR10+ version 1 SEI messages.

Encoder enhancements

  1. Create API function for allocating and freeing x265_analysis_data.
  2. CEA 608/708 support: Read SEI messages from text file and encode it using userSEI message.

Bug fixes

  1. Disable noise reduction when vbv is enabled.
  2. Support minLuma and maxLuma values changed by the commandline.

Version 2.8

Release date - 21/05/2018

New features

  1. --asm avx512 used to enable AVX-512 in x265. Default disabled.
    For 4K main10 high-quality encoding, we are seeing good gains; for other resolutions and presets, we don’t recommend using this setting for now.
  2. --dynamic-refine dynamically switches between different inter refine levels. Default disabled.
    It is recommended to use :option:`–refine-intra 4’ with dynamic refinement for a better trade-off between encode efficiency and performance than using static refinement.
  3. --single-sei
    Encode SEI messages in a single NAL unit instead of multiple NAL units. Default disabled.
  4. --max-ausize-factor controls the maximum AU size defined in HEVC specification.
    It represents the percentage of maximum AU size used. Default is 1.
  5. VMAF (Video Multi-Method Assessment Fusion) Added VMAF support for objective quality measurement of a video sequence. Enable cmake option ENABLE_LIBVMAF to report per frame and aggregate VMAF score. The frame level VMAF score does not include temporal scores. This is supported only on linux for now.

Encoder enhancements

  1. Introduced refine-intra level 4 to improve quality.
  2. Support for HLG-graded content and pic_struct in SEI message.

Bug Fixes

  1. Fix 32 bit build error (using CMAKE GUI) in Linux.
  2. Fix 32 bit build error for asm primitives.
  3. Fix build error on mac OS.
  4. Fix VBV Lookahead in analysis load to achieve target bitrate.

Version 2.7

Release date - 21st Feb, 2018.

New features

  1. --gop-lookahead can be used to extend the gop boundary(set by –keyint). The GOP will be extended, if a scene-cut frame is found within this many number of frames.
  2. Support for RADL pictures added in x265. --radl can be used to decide number of RADL pictures preceding the IDR picture.

Encoder enhancements

  1. Moved from YASM to NASM assembler. Supports NASM assembler version 2.13 and greater.
  2. Enable analysis save and load in a single run. Introduces two new cli options –analysis-save <filename> and –analysis-load <filename>.
  3. Comply to HDR10+ LLC specification.
  4. Reduced x265 build time by more than 50% by re-factoring ipfilter.asm.

Bug fixes

  1. Fixed inconsistent output issue in deblock filter and –const-vbv.
  2. Fixed Mac OS build warnings.
  3. Fixed inconsistency in pass-2 when weightp and cutree are enabled.
  4. Fixed deadlock issue due to dropping of BREF frames, while forcing slice types through qp file.

Version 2.6

Release date - 29th November, 2017.

New features

  1. x265 can now refine analysis from a previous HEVC encode (using options --refine-inter, and --refine-intra), or a previous AVC encode (using option --refine-mv-type). The previous encode’s information can be packaged using the x265_analysis_data_t data field available in the x265_picture object.
  2. Basic support for segmented (or chunked) encoding added with --vbv-end that can specify the status of CPB at the end of a segment. String this together with --vbv-init to encode a title as chunks while maintaining VBV compliance!
  3. --force-flush can be used to trigger a premature flush of the encoder. This option is beneficial when input is known to be bursty, and may be at a rate slower than the encoder.
  4. Experimental feature --lowpass-dct that uses truncated DCT for transformation.

Encoder enhancements

  1. Slice-parallel mode gets a significant boost in performance, particularly in low-latency mode.
  2. x265 now officially supported on VS2017.
  3. x265 now supports all depths from mono0 to mono16 for Y4M format.

API changes

  1. Options that modified PPS dynamically (--opt-qp-pps and --opt-ref-list-length-pps) are now disabled by default to enable users to save bits by not sending headers. If these options are enabled, headers have to be repeated for every GOP.
  2. Rate-control and analysis parameters can dynamically be reconfigured simultaneously via the x265_encoder_reconfig API.
  3. New API functions to extract intermediate information such as slice-type, scenecut information, reference frames, etc. are now available. This information may be beneficial to integrating applications that are attempting to perform content-adaptive encoding. Refer to documentation on x265_get_slicetype_poc_and_scenecut, and x265_get_ref_frame_list for more details and suggested usage.
  4. A new API to pass supplemental CTU information to x265 to influence analysis decisions has been added. Refer to documentation on x265_encoder_ctu_info for more details.

Bug fixes

  1. Bug fixes when --slices is used with VBV settings.
  2. Minor memory leak fixed for HDR10+ builds, and default x265 when pools option is specified.
  3. HDR10+ bug fix to remove dependence on poc counter to select meta-data information.

Version 2.5

Release date - 13th July, 2017.

Encoder enhancements

  1. Improved grain handling with --tune grain option by throttling VBV operations to limit QP jumps.
  2. Frame threads are now decided based on number of threads specified in the --pools, as opposed to the number of hardware threads available. The mapping was also adjusted to improve quality of the encodes with minimal impact to performance.
  3. CSV logging feature (enabled by --csv) is now part of the library; it was previously part of the x265 application. Applications that integrate libx265 can now extract frame level statistics for their encodes by exercising this option in the library.
  4. Globals that track min and max CU sizes, number of slices, and other parameters have now been moved into instance-specific variables. Consequently, applications that invoke multiple instances of x265 library are no longer restricted to use the same settings for these parameter options across the multiple instances.
  5. x265 can now generate a seprate library that exports the HDR10+ parsing API. Other libraries that wish to use this API may do so by linking against this library. Enable ENABLE_HDR10_PLUS in CMake options and build to generate this library.
  6. SEA motion search receives a 10% performance boost from AVX2 optimization of its kernels.
  7. The CSV log is now more elaborate with additional fields such as PU statistics, average-min-max luma and chroma values, etc. Refer to documentation of --csv for details of all fields.
  8. x86inc.asm cleaned-up for improved instruction handling.

API changes

  1. New API x265_encoder_ctu_info() introduced to specify suggested partition sizes for various CTUs in a frame. To be used in conjunction with --ctu-info to react to the specified partitions appropriately.
  2. Rate-control statistics passed through the x265_picture object for an incoming frame are now used by the encoder.
  3. Options to scale, reuse, and refine analysis for incoming analysis shared through the x265_analysis_data field in x265_picture for runs that use --analysis-reuse-mode load; use options --scale, --refine-mv, --refine-inter, and --refine-intra to explore.
  4. VBV now has a deterministic mode. Use --const-vbv to exercise.

Bug fixes

  1. Several fixes for HDR10+ parsing code including incompatibility with user-specific SEI, removal of warnings, linking issues in linux, etc.
  2. SEI messages for HDR10 repeated every keyint when HDR options (--hdr-opt, --master-display) specified.

Version 2.4

Release date - 22nd April, 2017.

Encoder enhancements

  1. HDR10+ supported. Dynamic metadata may be either supplied as a bitstream via the userSEI field of x265_picture, or as a json jile that can be parsed by x265 and inserted into the bitstream; use --dhdr10-info to specify json file name, and --dhdr10-opt to enable optimization of inserting tone-map information only at IDR frames, or when the tone map information changes.
  2. Lambda tables for 8, 10, and 12-bit encoding revised, resulting in significant enhancement to subjective visual quality.
  3. Enhanced HDR10 encoding with HDR-specific QP optimzations for chroma, and luma planes of WCG content enabled; use --hdr-opt to activate.
  4. Ability to accept analysis information from other previous encodes (that may or may not be x265), and selectively reuse and refine analysis for encoding subsequent passes enabled with the --refine-level option.
  5. Slow and veryslow presets receive a 20% speed boost at iso-quality by enabling the --limit-tu option.
  6. The bitrate target for x265 can now be dynamically reconfigured via the reconfigure API.
  7. Performance optimized SAO algorithm introduced via the --limit-sao option; seeing 10% speed benefits at faster presets.

API changes

  1. x265_reconfigure API now also accepts rate-control parameters for dynamic reconfiguration.
  2. Several additions to data fields in x265_analysis to support --refine-level: see x265.h for more details.

Bug fixes

  1. Avoid negative offsets in x265 lambda2 table with SAO enabled.
  2. Fix mingw32 build error.
  3. Seek now enabled for pipe input, in addition to file-based input
  4. Fix issue of statically linking core-utils not working in linux.
  5. Fix visual artifacts with --multi-pass-opt-distortion with VBV.
  6. Fix bufferFill stats reported in csv.

Version 2.3

Release date - 15th February, 2017.

Encoder enhancements

  1. New SSIM-based RD-cost computation for improved visual quality, and efficiency; use --ssim-rd to exercise.
  2. Multi-pass encoding can now share analysis information from prior passes (in addition to rate-control information) to improve performance and quality of subsequent passes; to your multi-pass command-lines that use the --pass option, add --multi-pass-opt-distortion to share distortion information, and --multi-pass-opt-analysis to share other analysis information.
  3. A dedicated thread pool for lookahead can now be specified with --lookahead-threads.
  4. option:–dynamic-rd dynamically increase analysis in areas where the bitrate is being capped by VBV; works for both CRF and ABR encodes with VBV settings.
  5. The number of bits used to signal the delta-QP can be optimized with the --opt-cu-delta-qp option; found to be useful in some scenarios for lower bitrate targets.
  6. Experimental feature option:–aq-motion adds new QP offsets based on relative motion of a block with respect to the movement of the frame.

API changes

  1. Reconfigure API now supports signalling new scaling lists.
  2. x265 application’s csv functionality now reports time (in milliseconds) taken to encode each frame.
  3. --strict-cbr enables stricter bitrate adherence by adding filler bits when achieved bitrate is lower than the target; earlier, it was only reacting when the achieved rate was higher.
  4. --hdr can be used to ensure that max-cll and max-fall values are always signaled (even if 0,0).

Bug fixes

  1. Fixed incorrect HW thread counting on MacOS platform.
  2. Fixed scaling lists support for 4:4:4 videos.
  3. Inconsistent output fix for --opt-qp-pss by removing last slice’s QP from cost calculation.
  4. VTune profiling (enabled using ENABLE_VTUNE CMake option) now also works with 2017 VTune builds.

Version 2.2

Release date - 26th December, 2016.

Encoder enhancements

  1. Enhancements to TU selection algorithm with early-outs for improved speed; use --limit-tu to exercise.
  2. New motion search method SEA (Successive Elimination Algorithm) supported now as :option: –me 4
  3. Bit-stream optimizations to improve fields in PPS and SPS for bit-rate savings through --opt-qp-pps, --opt-ref-list-length-pps, and --multi-pass-opt-rps.
  4. Enabled using VBV constraints when encoding without WPP.
  5. All param options dumped in SEI packet in bitstream when info selected.
  6. x265 now supports POWERPC-based systems. Several key functions also have optimized ALTIVEC kernels.

API changes

  1. Options to disable SEI and optional-VUI messages from bitstream made more descriptive.
  2. New option --scenecut-bias to enable controlling bias to mark scene-cuts via cli.
  3. Support mono and mono16 color spaces for y4m input.
  4. --min-cu-size of 64 no-longer supported for reasons of visual quality (was crashing earlier anyways.)
  5. API for CSV now expects version string for better integration of x265 into other applications.

Bug fixes

  1. Several fixes to slice-based encoding.
  2. --log2-max-poc-lsb’s range limited according to HEVC spec.
  3. Restrict MVs to within legal boundaries when encoding.

Version 2.1

Release date - 27th September, 2016

Encoder enhancements

  1. Support for qg-size of 8
  2. Support for inserting non-IDR I-frames at scenecuts and when running with settings for fixed-GOP (min-keyint = max-keyint)
  3. Experimental support for slice-parallelism.

API changes

  1. Encode user-define SEI messages passed in through x265_picture object.
  2. Disable SEI and VUI messages from the bitstream
  3. Specify qpmin and qpmax
  4. Control number of bits to encode POC.

Bug fixes

  1. QP fluctuation fix for first B-frame in mini-GOP for 2-pass encoding with tune-grain.
  2. Assembly fix for crashes in 32-bit from dct_sse4.
  3. Threadpool creation fix in windows platform.

Version 2.0

Release date - 13th July, 2016

New Features

  1. uhd-bd: Enable Ultra-HD Bluray support
  2. rskip: Enables skipping recursion to analyze lower CU sizes using heuristics at different rd-levels. Provides good visual quality gains at the highest quality presets.
  3. rc-grain: Enables a new ratecontrol mode specifically for grainy content. Strictly prevents QP oscillations within and between frames to avoid grain fluctuations.
  4. tune grain: A fully refactored and improved option to encode film grain content including QP control as well as analysis options.
  5. asm: ARM assembly is now enabled by default, native or cross compiled builds supported on armv6 and later systems.

API and Key Behaviour Changes

  1. x265_rc_stats added to x265_picture, containing all RC decision points for that frame
  2. PTL: high tier is now allowed by default, chosen only if necessary
  3. multi-pass: First pass now uses slow-firstpass by default, enabling better RC decisions in future passes
  4. pools: fix behaviour on multi-socketed Windows systems, provide more flexibility in determining thread and pool counts
  5. ABR: improve bits allocation in the first few frames, abr reset, vbv and cutree improved

Misc

  1. An SSIM calculation bug was corrected

Version 1.9

Release date - 29th January, 2016

New Features

  1. Quant offsets: This feature allows block level quantization offsets to be specified for every frame. An API-only feature.
  2. –intra-refresh: Keyframes can be replaced by a moving column of intra blocks in non-keyframes.
  3. –limit-modes: Intelligently restricts mode analysis.
  4. –max-luma and –min-luma for luma clipping, optional for HDR use-cases
  5. Emergency denoising is now enabled by default in very low bitrate, VBV encodes

API Changes

  1. x265_frame_stats returns many additional fields: maxCLL, maxFALL, residual energy, scenecut and latency logging
  2. –qpfile now supports frametype ‘K”
  3. x265 now allows CRF ratecontrol in pass N (N greater than or equal to 2)
  4. Chroma subsampling format YUV 4:0:0 is now fully supported and tested

Presets and Performance

  1. Recently added features lookahead-slices, limit-modes, limit-refs have been enabled by default for applicable presets.
  2. The default psy-rd strength has been increased to 2.0
  3. Multi-socket machines now use a single pool of threads that can work cross-socket.

Version 1.8

Release date - 10th August, 2015

API Changes

  1. Experimental support for Main12 is now enabled. Partial assembly support exists.
  2. Main12 and Intra/Still picture profiles are now supported. Still picture profile is detected based on x265_param::totalFrames.

3. Three classes of encoding statistics are now available through the API. a) x265_stats - contains encoding statistics, available through x265_encoder_get_stats() b) x265_frame_stats and x265_cu_stats - contains frame encoding statistics, available through recon x265_picture 4. –csv a) x265_encoder_log() is now deprecated b) x265_param::csvfn is also deprecated 5. –log-level now controls only console logging, frame level console logging has been removed. 6. Support added for new color transfer characteristic ARIB STD-B67

New Features

  1. limit-refs: This feature limits the references analysed for individual CUS. Provides a nice tradeoff between efficiency and performance.
  2. aq-mode 3: A new aq-mode that provides additional biasing for low-light conditions.
  3. An improved scene cut detection logic that allows ratecontrol to manage visual quality at fade-ins and fade-outs better.

Preset and Tune Options

  1. tune grain: Increases psyRdoq strength to 10.0, and rdoq-level to 2.
  2. qg-size: Default value changed to 32.