encoding.info

Settings index

Encoder options and the pages that cover them.

SettingWhat it doesExplained on
--ampx265Lets a coding unit be predicted as two unequal parts, one a quarter of the unit and the other three quarters (asymmetric motion partitions). Off in the medium preset, on from slower.
--aq-modeSVT-AV1Adaptive quantization mode, 2 by default. Only the name, default and documentation were read.Not yet checked against source.
--aq-modeaomenc (libaom)Adaptive quantization mode, 0 (off) by default. Only the name, help text and default were read.Not yet checked against source.
--aq-modex264Chooses how block energy becomes a per-block quantization parameter (QP) offset: 0 is off, 1 uses a fixed zero point, 2 scales with the frame's statistics, 3 adds extra bits for low-energy blocks. 1 in every preset except ultrafast, which turns it off.
--aq-modex265Chooses how block energy becomes a per-block quantization parameter (QP) offset: 0 is off, 1 uses a fixed zero point, 2 takes its reference from the frame's statistics, 3 adds a bias that lowers QP on low-energy blocks, 4 is mode 2 with edge density standing in for energy on blocks that contain edges. 2 in the medium preset; ultrafast and superfast turn it off.
--aq-strengthx264Scales every adaptive quantization offset. Default 1.0; 0 turns the offsets off.
--audx264Writes an access unit delimiter at the start of each frame. Off by default.
--audx265Writes access unit delimiters. Off by default.
--b-adaptx264How the lookahead places B frames: 0 uses a fixed pattern (B frames up to the limit, then a P frame, cut short at keyframes and at the end of the video), 1 (the default) decides quickly, 2 searches over patterns and is used by the slower presets.
--b-pyramidx264Makes the middle B frame of a run of two or more a reference for the others: none, strict or normal (the default). strict marks the reference B frame at a lower priority.
--b-pyramidx265On by default; x265's structure was not read.Not yet checked against source.
--bframesx264The most B frames x264 places in a row, 3 by default; the lookahead decides how many to use at each point. 0 in the ultrafast preset, the zerolatency tuning and the Baseline profile; 8 in veryslow, 16 in placebo; the animation tuning adds 2.
--bframesx265The most B frames in a row, 4 by default.
--bitratex264Average bitrate in kbit/s, reached by feedback in one pass or by planning in two.
cabac-idcx264A library parameter, not listed among the command-line options: picks which of three tables of starting states the arithmetic coder's contexts use in slices that are not intra. 0 to 2, default 0.
--chromalocx264Labels where 4:2:0 chroma samples sit relative to luma. Written to the stream only for 4:2:0 and values 1 to 5.
--colormatrixx264Writes a label into the stream saying which weights turn luma and chroma back into red, green and blue. It converts nothing.
--cplxblurx264How far the second pass smooths frame complexity over neighboring frames; 20 by default.
--cpu-usedaomenc (libaom)Speed setting; a higher number is faster. The library's default is 0. Only the name, help text and default were read.Not yet checked against source.
--cq-levelaomenc (libaom)The quality level for the cq and q modes, 10 by default. The library starts in vbr mode, so the 10 is stored and not used until --end-usage selects one of those modes. Only the name and default were read.Not yet checked against source.
--cqmx264Chooses the quantization matrices: flat (every weight 16, the default) or jvt, the JVT matrix set, whose 8x8 intra matrix runs from 6 at the lowest frequency to 42 at the highest.
--cqmfilex264Reads custom quantization matrices from a file and overrides the other matrix options.
--crfSVT-AV1Constant rate factor, 35 by default per its documentation. Only the name and documentation were read.Not yet checked against source.
--crfx264Constant rate factor, 23 by default and the default mode. It sets a rate factor, not a bitrate or a measured quality; x264 scales it so the number is somewhat comparable with a QP, and each frame's quantizer moves with complexity as --qcomp allows. Not usable with two passes.
--ctux265The size of the largest block, the coding tree unit: 16, 32 or 64. 64 by default, 32 in ultrafast and superfast.
--cutreex265Coding-unit tree, on by default. Its code was not read.Not yet checked against source.
--deadzone-interx264Sets the nominal rounding offset for luma in inter macroblocks to (32 - value) / 64, so a larger value widens the band of coefficients that quantize to zero. Default 21, range 0 to 32. Affects only the fixed-offset path, not blocks where trellis quantization runs.
--deadzone-intrax264Sets the nominal rounding offset for luma in intra macroblocks to (32 - value) / 64, so a larger value widens the band of coefficients that quantize to zero. Default 11, range 0 to 32. Affects only the fixed-offset path, not blocks where trellis quantization runs.
--deblockx264Two offsets, alpha:beta, each from -6 to 6 and 0:0 by default; one number sets both. Each unit moves the deblocking filter's threshold tables by two steps of QP: positive filters more, negative less. The film, grain, stillimage, touhou and animation tunings change it.
--deblockx265Two offsets, tC:beta, each from -6 to 6 and 0:0 by default; one number sets both. The animation tuning uses 1:1. The filter's arithmetic was not read in x265.
--deltaq-modeaomenc (libaom)A second per-block quantizer control, mode 1 by default. Only the name, help text and default were read.Not yet checked against source.
--denoise-noise-levelaomenc (libaom)0 to 50. The option list describes it as the amount of noise to remove before encoding, with grain parameters estimated for synthesis; the code was not read. The option exists only in builds with denoising, and --enable-dnl-denoising, on by default in the option list, decides whether the frame being coded is denoised.Not yet checked against source.
--ditherx265Dithers an input that has more bits than the encoder's depth. Off unless given.
--dump-yuvx264Writes the encoder's reconstructed frames, after deblocking, to a file: the pictures a decoder will produce.
--enable-cdefaomenc (libaom)Takes 0 to 3. The option list describes them as off, on (the default), off for frames that are not references, and chosen from the frame's quantizer; the code behind them was not read.Not yet checked against source.
--enable-qmaomenc (libaom)Turns AV1 quantization matrices on. Off by default.
--enable-restorationaomenc (libaom)Loop restoration, 0 or 1. The option list gives 1 as the default except in realtime mode; the code behind it was not read.Not yet checked against source.
--end-usageaomenc (libaom)Rate control mode: vbr (the library's default), cbr, cq or q. Only the name and default were read.Not yet checked against source.
--film-grainSVT-AV1Film grain level, 0 to 50, off by default. A separate option, --film-grain-denoise, off by default, decides whether the coded pictures are denoised. Only the names, defaults and documentation were read.Not yet checked against source.
--film-grain-tableaomenc (libaom)The path of a file of film grain parameters for the decoder to synthesize from; described from the option list.Not yet checked against source.
--input-cspx265Sets the chroma format the encoder works in, 4:2:0 unless given. Any other format gets a range extensions profile.
--ipratiox264An I frame's quantizer scale is a running average of recent P frame levels divided by this; 1.4 by default, about 2.9 QP lower.
--keyintSVT-AV1Keyframe interval; the default of -2 means about five seconds per its documentation. Only the name, default and documentation were read.Not yet checked against source.
--keyintx264The longest distance between keyframes, 250 frames by default. The same default in x265.
--kf-max-distaomenc (libaom)Longest keyframe interval in frames, 9999 by default. Only the name and default were read.Not yet checked against source.
--lag-in-framesaomenc (libaom)How many frames the encoder may hold before output, 35 by default. Only the name and default were read.Not yet checked against source.
--lambda-filex265Loads replacement lambda tables in place of the built-in ones. The option was found in source; its file format was not read.Not yet checked against source.
--levelx264States the level. By default x264 takes the first row of its level table that the stream's size, rate and buffer settings pass, and it warns when a requested level is exceeded.
--level-idcx265States the level. By default x265 picks the lowest level, and tier, whose limits the picture size, frame rate and bitrate fit.
--limit-saox265Has x265 try fewer offset types in B slices and for skipped blocks in P slices. Off by default.
--lookaheadSVT-AV1Frames to look ahead, beyond what its frame grouping, temporal filtering and rate control already need; -1, the default, is automatic. It is not a measure of total delay. Only the name, default and documentation were read.Not yet checked against source.
--max-tu-sizex265Largest transform size allowed: 32 (default), 16, 8 or 4.
--mex264The whole-pixel motion search pattern: dia, hex, umh, esa or tesa. hex in the medium preset, dia in ultrafast and superfast, umh in slower and veryslow, tesa in placebo.
--merangex264The extent of the whole-pixel motion search; what that means depends on the search pattern. 16 by default, 24 in veryslow and placebo. Accepted range 4 to 1024; for dia and hex, values above 16 are reduced to 16.
--min-cu-sizex265The smallest coding unit the tree may be split down to: 8, 16 or 32. 8 by default, 16 in ultrafast.
--min-keyintx264The shortest distance at which x264 turns an I frame it has placed into a keyframe. By default the smaller of a tenth of --keyint and the frame rate. An I frame closer than this to the last keyframe, for instance at a scene cut, stays a plain I frame. A frame given to x264 as IDR is a keyframe at any distance.
--mixed-refsx264Lets each 8x8 quarter of a macroblock use a different reference frame; without it the quarters reuse the frame chosen for the whole macroblock. On in the medium preset, off in ultrafast to faster, and without effect when --ref is 1.
--no-8x8dctx264Removes the 8x8 transform, leaving 4x4 only. By default the encoder chooses between the two for each macroblock.
--no-cabacx264Switches from the arithmetic coder to the variable-length coder. The arithmetic coder is on by default; the ultrafast preset, the fastdecode tuning and the Baseline profile turn it off.
--no-deblockx264Turns the deblocking filter off: x264 writes into each slice header that the filter is off, so the decoder does not run it either. The ultrafast preset and the fastdecode tuning do the same.
--no-deblockx265Turns x265's deblocking filter off. The fast-decode tuning does the same.
--no-fast-pskipx264Turns off the early shortcut that can end a P macroblock's search with skip. With the shortcut on and --subme 3 or more, a 16x16 search on the first reference with a small prediction cost and a vector close to the predicted skip vector triggers a skip probe, and x264 takes skip if the probe succeeds. Skip can still be chosen without the shortcut. Fast skip is on by default; only placebo turns it off.
--no-high-tierx265Forbids the High tier, so a bitrate or buffer size over a level's Main limit moves the stream to the next level instead. The High tier is allowed by default.
--no-mbtreex264Turns macroblock-tree off. It is on by default and off in the ultrafast and superfast presets and the zerolatency tuning. With it on, --qcomp sets its strength.
--no-psyx264Turns both psychovisual terms off. The psnr and ssim tunings do the same. It also changes how chroma is weighted.
--open-gopx264Lets B frames just before a keyframe use it and the previous group's last P frame; the keyframe is then an I frame with a recovery point, not an IDR frame. Off by default.
--open-gopx265On by default in x265.
--output-cspx264Chooses the chroma format the encoder works in: i420 unless set, or i422 or i444. The program converts the input to it, and the format decides the profile.
--output-depthx264Sets the bit depth the encoder works in, 8 or 10. The program dithers an input that has more bits.
--output-depthx265Sets the bit depth the encoder works in. It has to match the depth the library was built for.
--partitionsx264Which block sizes are tried. p8x8 allows a macroblock in a predicted frame to be halved either way or quartered; p4x4 allows each quarter to be divided again and needs p8x8; b8x8 is the same for bidirectional frames; i8x8 and i4x4 allow the smaller intra blocks, and i8x8 needs the 8x8 transform. Medium preset: p8x8,b8x8,i8x8,i4x4. p4x4 is added from the slower preset; ultrafast switches all of them off.
--passx2641 writes a statistics file from a first encode; 2 reads it and plans the second. The first pass uses faster settings unless --slow-firstpass is given.
--pbratiox264A B frame's QP is its neighbors' average plus 6 x log2 of this; 1.3 by default, about 2.3 QP higher. A reference B frame gets half the offset. With macroblock-tree on, the default, x264 sets this ratio to 1 and the offset is not applied.
--presetSVT-AV1Speed setting up to 13, 8 by default; a higher number is faster. Only the name, default and documentation were read.Not yet checked against source.
--presetx264Selects a bundle of values for the twenty grouped settings on the presets page, from ultrafast to placebo; medium leaves the defaults. On x264's command line, options given with it override individual values, and profile restrictions and x264's own validation can change what takes effect.
--profilex264Forces the limits of an H.264 profile by switching off the tools it lacks: baseline removes the 8x8 transform, arithmetic coding, B-frames and weighted prediction. It converts nothing: a request the profile cannot hold (lossless coding, 4:4:4, 4:2:2, more than 8 bits, monochrome) stops x264 with an error. Without it x264 names the lowest profile that covers the tools in use.
--psnrx264Reports peak signal-to-noise ratio, computed from the squared error between the encoder's reconstruction and the source: per frame, as a mean of the frame values for each plane, and as a global value from the pooled squared error of all planes, each frame weighted by its duration. Off by default.
--psy-rdx264Two strengths of psychovisual optimization. The first adds a cost in mode decision for how far the reconstruction's texture energy is from the source's, and needs --subme 6 or higher. The second biases trellis quantization toward keeping coefficient energy and needs trellis; default 0.
--qblurx264Smoothing applied to the second pass's quantizers; 0.5 by default.
--qcompx264Quantizer compression, 0.6 by default. With macroblock-tree off: at 1 the base quantizer stops following a frame's complexity (frame types still differ), and at 0 frames of one type come out nearly the same size. With macroblock-tree on, the default, it sets macroblock-tree's strength instead, and a value of 1 turns macroblock-tree off.
--qm-maxaomenc (libaom)Highest (flattest) quantization matrix level the encoder may use, on AV1's scale of 0 to 15 where 15 is flat. Default 9.
--qm-minaomenc (libaom)Lowest (steepest) quantization matrix level the encoder may use, on AV1's scale of 0 to 15 where 15 is flat. Default 5.
--qpx264Constant quantizer: the QP of every P frame. I and B frames get it minus and plus the offsets from --ipratio and --pbratio.
--qpstepx264The usual bound on how far QP moves from one frame to the next of the same type in one-pass average bitrate; 4 by default. Strong feedback widens the bound and the buffer rule acts after it.
--rangex264Labels the stream's values as full range or the narrower video range. The x264 program also converts the input when its range differs.
--ratetolx264How strongly one-pass average bitrate feedback reacts to drift from its target; 1.0 by default, and a larger value reacts more gently. It is not a band without feedback or a guarantee around the target.
--rcSVT-AV1Rate control mode: 0 (the default) is constant rate factor, or constant quantizer when --aq-mode is 0; 1 is variable bitrate; 2 is constant bitrate. Only the name, default and documentation were read.Not yet checked against source.
--rc-lookaheadx264How far ahead x264's rate control and frame planning look: 40 frames by default, 0 in ultrafast, superfast and the zerolatency tuning, 60 in slower and above. It is not the total delay: it counts toward x264's input delay only when macroblock-tree or a buffer limit is on, and B frames and threads add their own. At 0 the scene-cut test and the B frame decisions still run.
--rdoq-levelx265Rate-distortion optimized quantization, x265's counterpart of trellis. 0 uses fixed rounding offsets with no cost search; 1 chooses levels by rate-distortion cost; 2 also considers zeroing whole groups of coefficients and dropping trailing levels of 1. 0 in the medium preset; the slow preset and those slower use 2.
--rectx265Lets a coding unit be predicted as two equal halves, side by side or one above the other. Off in the medium preset, on from slow.
--refx264The most reference frames a frame may choose from, 1 to 16. 3 in the medium preset; 1 in ultrafast to veryfast, 2 in faster and fast, 5 in slow, 8 in slower, 16 in veryslow and placebo.
--refx265The most reference frames a frame may choose from. 1 to 16 is accepted, but the count plus the number of reordered pictures must be at most 8 to fit any level: with a level requested x265 lowers it to fit (8, or 7 with B-frames), and without one it marks the stream as outside every level. 3 in the medium preset; 1 in ultrafast and superfast, 2 in veryfast and faster, 4 in slow, 5 from slower.
--repeat-headersx265Writes the parameter sets with every keyframe. Off by default.
--saox265Sample adaptive offset: on by default, off with --no-sao, in the ultrafast and superfast presets and in the grain and fast-decode tunings.
--sao-non-deblockx265The option exists and is off by default; what it changes was not read in the code.Not yet checked against source.
--scaling-listx265Quantization matrices (scaling lists): off (the default, flat weighting), default (HEVC's standard set, switched on by a flag without sending the values), or a file of custom lists, whose values are sent.
--scenecutx264How readily the lookahead calls a frame a scene cut and codes it as intra: 40 by default, and 0 or --no-scenecut turns the test off. A frame is a cut when its estimated P cost reaches (1 - bias) times its I cost, where the bias grows from this value and the distance from the last keyframe. x265 has the same default.
--selective-saox265Takes 0 to 4, default 0; what the values select was not read in the code.Not yet checked against source.
--ssimx264Reports structural similarity of the luma plane, computed over 8x8 windows every 4 pixels, as a mean and in decibels. Off by default.
--statsx264The name of the statistics file for two-pass encoding.
--submex264Sets how thorough sub-pixel motion search and mode decision are. From 2 upward, sub-pixel and fast mode comparisons use SATD instead of SAD. Reconstruction-based mode decision starts at 6 in I and P slices and at 7 in B slices. At 0 motion vectors stay on whole pixels; at 1 only the partition finally chosen is refined; at 2 and 3 every search gets a half-pixel round and the quarter-pixel rounds are left for the chosen partition; from 4 every search gets quarter-pixel rounds too. 7 in the medium preset. Levels 10 and 11 need --trellis 2 and adaptive quantization; without them x264 lowers the level to 9.
--trellisx264Trellis quantization: 0 is off, 1 runs it on the final encode of each macroblock, 2 also runs it inside mode decision. 1 in the medium preset; the slow preset and those slower use 2. When it runs, the dead zone settings are not used.
--tu-inter-depthx265How many levels of transform quadtree splitting are searched for inter blocks, 1 to 4. Default 1; slower presets raise it.
--tu-intra-depthx265How many levels of transform quadtree splitting are searched for intra blocks, 1 to 4. Default 1; slower presets raise it.
--tuneSVT-AV1What the encoder tunes for, one of six modes in its documentation; 1, the default, is PSNR. Only the name, default and documentation were read.Not yet checked against source.
--tuneaomenc (libaom)The metric the encoder tunes for, psnr by default. Only the name and default were read.Not yet checked against source.
--tune grainx264Among several other changes, sets --deadzone-inter and --deadzone-intra to 6, narrowing the zero band in the fixed-offset path so that more small coefficients survive. Its other changes are not yet covered.
--tune psnrx264Turns off adaptive quantization and psychovisual optimizations, to maximize a mean-squared-error score rather than how the picture looks.
--tune ssimx264Selects aq-mode 2 and turns off the psychovisual optimizations.
--tune zerolatencyx264Sets the lookahead, the threaded lookahead and B frames to zero, switches macroblock-tree off, uses sliced threads and treats the input as constant frame rate, so coding a frame does not wait on those tools. It does not set the delay of the whole delivery chain.
--vbv-bufsizex264The size of the buffer model, in kilobits. Unset by default, so no buffer is modeled.
--vbv-initx264The buffer's starting level, 0.9 of its size by default.
--vbv-maxratex264The fill rate of the buffer model, in kbit/s. Unset by default. With a buffer size, in one-pass average bitrate, a value at or below --bitrate turns on x264's minimum-rate behavior, the usual constant bitrate setup; filler is added only with --filler or --nal-hrd cbr.