X264 Settings

From MeWiki

Jump to: navigation, search

Contents

[hide]

This page explains the purpose and use of all x264 parameters. Parameters are ordered the same as their appearance in x264 --fullhelp.

See also x264 Stats Output, x264 Stats File and x264 Encoding Suggestions.

[edit] x264 Settings

[edit] Help

x264 comes with some inbuilt documentation. To access this help, run x264 with --help, --longhelp or --fullhelp. Each successive option will give more detailed information.

[edit] Input

Specify the input video with a single positional argument. For example: 

x264.exe --output NUL C:\input.avs
x264 --output /dev/null ~/input.y4m

When your input is raw YUV you must also tell x264 the input resolution. You probably want to specify the framerate with --fps as well: 

x264.exe --output NUL --fps 25 --input-res 1280x720 D:\input.yuv 
x264 --output /dev/null --fps 30000/1001 --input-res 640x480 ~/input.yuv

[edit] Presets

A system designed to reduce the work needed to generate sane, efficient commandlines that do what you want. For information on what options these settings toggle, check the output of x264.exe --fullhelp.

[edit] profile

Default: not set

Limit the profile of the output stream. If you specify a profile, it overrides all other settings, so if you use it, you will be guaranteed a compatible stream. If you set this option, you cannot use lossless encoding (--qp 0 or --crf 0).

You should set this if you know your playback device only supports a certain profile. Most decoders support High profile, so there's no need to set this.

Values available: baseline, main, high.

[edit] preset

Default: medium

Change options to trade off compression efficiency against encoding speed. If you specify a preset, the changes it makes will be applied before all other parameters are applied.

You should generally set this option to the slowest you can bear.

Values available: ultrafast, superfast, veryfast, faster, fast, medium, slow, slower, veryslow, placebo.

[edit] tune

Default: Not Set

Tune options to further optimize them for your input content. If you specify a tuning, the changes will be applied after --preset but before all other parameters.

If your source content matches one of the available tunings you can use this, otherwise leave unset.

Values available: film, animation, grain, stillimage, psnr, ssim, fastdecode, zerolatency.

[edit] slow-firstpass

Default: Not Set

Using --pass 1 applies the following settings at the end of parsing the command line:

You can set --slow-firstpass to disable the feature, set this option. Note that using --preset placebo enables slow-firstpass too.

See --pass.

[edit] Frame-type options

[edit] keyint

Default: 250

Sets the maximum interval between IDR-frames (aka keyframes) in x264's output. You can specify "infinite" to never insert non-scenecut IDR-frames.

IDR-frames are 'delimiters' in the stream - no frame can reference data from the other side of the IDR-frame. As well as this, IDR-frames are also I-frames, so they don't reference data from any other frame. This means they can be used as seek points in a video.

Note that I-frames are generally significantly larger than P/B-frames (often 10x or more in low motion scenes), so they can play havoc with ratecontrol when combined with aggressively low VBV settings (eg, sub-second buffer sizes). In these cases, investigate --intra-refresh.

The default setting is fine for most videos. When encoding for Blu-ray, broadcast, live streaming or certain other specialist scenarios you may require a significantly smaller GOP length (often ~1x fps).

See also: --min-keyint, --scenecut, --intra-refresh

[edit] min-keyint

Default: auto (MIN(--keyint / 10,--fps))

Sets the minimum length between IDR-frames.

See --keyint for an explanation of IDR-frames. Very small keyint ranges can cause "incorrect" IDR-frame placement (for example, a strobing scene). This option limits the minimum length in frames after each IDR-frame before another can be placed.

The maximum allowed value for min-keyint is --keyint/2+1

Recommendation: Default, or 1x your framerate.

See also: --keyint, --scenecut

[edit] no-scenecut

Default: Not Set

Completely disables adaptive I-frame decision.

See also: --scenecut

[edit] scenecut

Default: 40

Sets the threshold for I/IDR frame placement (read: scene change detection).

x264 calculates a metric for every frame to estimate how different it is from the previous frame. If the value is lower than scenecut, a 'scenecut' is detected. An I-frame is placed if it has been less than --min-keyint frames since the last IDR-frame, otherwise an IDR-frame is placed. Higher values of scenecut increase the number of scenecuts detected. For more information on how the scenecut comparison works, see this doom9 thread.

Setting scenecut to 0 is equivalent to setting --no-scenecut.

Recommendation: Default

See also: --keyint, --min-keyint, --no-scenecut

[edit] intra-refresh

Default: Off

Disables IDR-frames, instead x264 uses intra coding for each macroblock in the frame every --keyint frames. Blocks are refreshed in a horizontally scrolling column - the 'refresh wave'. This benefits low-latency streaming by making it possible to achieve more constant frame sizes than is possible with standard IDR-frames. It also increases the resilience of the video stream to packet loss. This option does reduce compression efficiency, hence only use it when needed.

Fun facts:

  • The first frame is still an IDR frame.
  • Intra-blocks are placed only in P-frames - the refresh wave is wider in the first P-frame after one or more B-frames.
  • The loss in compression efficiency comes primarily from the fact macroblocks on the 'new' (left) side of the refresh wave can't refer to data on the 'old' (right) side.

[edit] bframes

Default: 3

Sets the maximum number of concurrent B-frames that x264 can use.

Without B-frames, a typical x264 stream has frame types like so: IPPPPP...PI. With --bframes 2, up to two consecutive P-frames can be replaced with B-frames, like: IBPBBPBPPPB...PI.

B-frames are similar to P-frames, except they can use motion prediction from future frames as well. This can lead to significantly better efficiency in terms of compression ratio. Their average quality is controlled by --pbratio.

Fun facts:

  • x264 also occasionally distinguishes between two different sorts of B-frames. A 'B' can refer to a B-frame that is used as a reference by other frames (see --b-pyramid, while 'b' can refer to a B-frame that is not. If you see a mix of 'B' and 'b', it's usually related to the above. When the difference is not important, just 'B' is generally used to refer to all B-frames.
  • For more information about the choices x264 makes between P or B frames for each candidate frame, see this ffmpeg-devel mail. In this case, frame types will look like this (in display order) (assuming --bframes 3): IBBBPBBBPBPI.

See also: --no-b-adapt, --b-bias, --b-pyramid, --ref, --pbratio, --partitions, --weightb

[edit] b-adapt

Default: 1

Set the adaptive B-frame placement decision algorithm. This setting controls how x264 decides between placing a P- or B-frame.

0. Disabled. Pick B-frames always. This is the same as what the older no-b-adapt setting did.
1. 'Fast' algorithm, faster, speed slightly increases with higher --b-frames setting. When using this mode, you basically always want to use --bframes 16.
2. 'Optimal' algorithm, slower, speed significantly decreases with higher --b-frames setting.

Note: For a multi-pass encode, this option is only needed for the first pass where frame types are decided.

[edit] b-bias

Default: 0

Controls the likelihood of B-frames being used instead of P-frames. Values greater than 0 increase the weighting towards B-frames, while values less than 0 do the opposite. This number is an arbitrary metric. The range is from -100 to 100. A value of 100/-100 does not guarantee every/no P-frame will be converted (use --b-adapt 0 for that).

Only use this if you think you make better ratecontrol decisions than x264.

See also: --b-frames, --ipratio

[edit] b-pyramid

Default: normal

Allow the use of B-frames as references for other frames. Without this setting, frames can only reference I- or P-frames. Although I/P-frames are more valued as references because of their higher quality, B-frames can also be useful. B-frames designated as references will get a quantizer halfway between P-frames and normal B-frames. You need to use at least two B-frames before B-pyramid will work.

If you're encoding for Blu-ray, use 'none' or 'strict'.

  • none: do not allow B-frames to be used as references.
  • strict: allow one B-frame per minigop to be used as reference; enforces restrictions imposed by the Blu-ray standard.
  • normal: allow numerous B-frames per minigop to be used as references.

See also: --bframes, --refs, --no-mixed-refs

[edit] open-gop

Default: none

Open-GOP is an encoding technique which increases efficiency. There are three modes:

  • none: Open-GOP disabled.
  • normal: Open-GOP enabled.
  • bluray: . Open-GOP enabled. A less efficient version of open-GOP, normal mode doesn't work when authoring Blu-rays.

Some decoders don't fully support open-GOP streams, which is why this hasn't been enabled by default. You should test with all decoders your streams will be played on, or (if that's impossible) wait until support is generally available.

There's an explanation of Open-GOP here.

[edit] no-cabac

Default: Not set

Disables CABAC (Context Adaptive Binary Arithmetic Coder) stream compression and falls back to the less efficient CAVLC (Context Adaptive Variable Length Coder) system. Significantly reduces both the compression efficiency (10-20% typically) and the decoding requirements.

[edit] ref

Default: 3

Controls the size of the DPB (Decoded Picture Buffer). The range is from 0-16. In short, this value is the number of previous frames each P-frame can use as references. (B-frames can use one or two fewer, depending on if they are used as references or not.) The minimum number of refs that can be referenced is 1.

Also note that the H.264 spec limits DPB size for each level. If adhering to Level 4.1 specs, the maximum refs for 720p and 1080p video are 9 and 4 respectively. You can read more about levels and 4.1 in particular under --level.

See also: --b-pyramid, --no-mixed-refs, --level

[edit] no-deblock

Default: Not Set

Completely disables the loop filter. Not Recommended.

See Also: --deblock

[edit] deblock

Default: 0:0

Controls the loop filter (aka inloop deblocker), which is part of the H.264 standard. It is very efficient in terms of encoding time vs. quality gained.

You can find a good description of how the loop filter parameters work in this doom9 thread (see the initial post and akupenguin's replies).

See Also: --no-deblock

[edit] slices

Default: 0

Sets the number of slices per frame, and forces rectangular slices. (Overridden by either --slice-max-size or --slice-max-mbs if they are set.)

If you are encoding for Blu-ray, set this to four. Otherwise, don't use this unless you know you need to.

See Also: --slice-max-size, --slice-max-mbs.

[edit] slice-max-size

Default: 0

Sets the maximum slice size in bytes, including estimated NAL overhead. (Currently is not compatible with --interlaced.)

See Also: --slices

[edit] slice-max-mbs

Default: 0

Sets the maximum slice size in macroblocks. (Currently is not compatible with --interlaced.)

See Also: --slices

[edit] tff

Enable interlaced encoding and specify the top field is first. x264's interlaced encoding uses MBAFF, and is inherently less efficient than progressive encoding. For that reason, you should only encode interlaced if you intend to display the video on an interlaced display (or can't deinterlace the video before sending it to x264). Implies --pic-struct.

[edit] bff

Enable interlaced encoding and specify the bottom field is first. See --tff for more info.

[edit] constrained-intra

Default: Not Set

Enable constrained intra prediction, which is required for the base layer of SVC encodes. Since EveryoneTM ignores SVC you can likewise ignore this switch.

[edit] pulldown

Default: none

Signal soft telecine for your (progressive, constant framerate) input stream using one of a few preset modes. Soft telecine is explained in more detail on the HandBrake wiki. The available presets are: none, 22, 32, 64, double, triple and euro.

Specifying any mode but none implies --pic-struct.

[edit] fake-interlaced

Default: Not Set

Mark a stream as interlaced even when not encoding as interlaced. Allows encoding of 25p and 30p Blu-ray compliant videos.

[edit] Ratecontrol

[edit] qp

Default: Not Set

The first of three possible ratecontrol methods. Set x264 to encode the movie in Constant Quantizer mode. The number you give here specifies the P-frame quantizer. The quantizer used for I- and B-frames is derived from --ipratio and --pbratio. CQ mode targets a certain quantizer, which means final filesize is not known (although it can be reasonably accurately estimated with some methods). A setting of 0 will produce lossless output. qp produces larger files than --crf for the same visual quality. qp mode also disables adaptive quantization, since by definition 'constant quantizer' implies no adaptive quantization.

This option is mutually exclusive with --bitrate and --crf. See this writeup for more information on the various ratecontrol systems.

You should generally use --crf instead, although qp doesn't require lookahead to run and thus can be faster.

See also: --bitrate, --crf, --ipratio, --pbratio

[edit] bitrate

Default: Not Set

The second of three ratecontrol methods. Encode the video in target bitrate mode. Target bitrate mode means the final filesize is known, but the final quality is not. x264 will attempt to encode the video to target the given bitrate as the overall average. The parameter given is the bitrate in kilobits/sec. (8bits = 1byte and so on). Note that 1 kilobit is 1000, not 1024 bits.

This setting is often used in conjunction with --pass for two-pass encoding.

This option is mutually exclusive with --qp and --crf. See this writeup for more information on the various ratecontrol systems.

See also: --qp, --crf, --ratetol, --pass, --stats

[edit] crf

Default: 23.0

The final ratecontrol method: Constant Ratefactor. While qp targets a certain quantizer, and bitrate targets a certain filesize, crf targets a certain 'quality'. The idea is for crf n to give the same perceptual quality as qp n, just in a smaller space. The arbitrary unit of measure for crf values is the "ratefactor".

CRF achieves this by reducing the quality of 'less important' frames. In this context, 'less important' means frames in complex or high-motion scenes, where quality is either more expensive (in terms of bits) or less visible, will have their quantizer increased. The bits saved in frames like these are redistributed to frames where they will be more effective.

CRF will take less time than a 2pass bitrate encode, because the 'first pass' from a 2pass encode was skipped. On the other hand, it's impossible to predict the bitrate a CRF encode will come out to. It's up to you to decide which rate-control mode is better for your circumstances.

This option is mutually exclusive with qp and bitrate. See this writeup for more information on the various ratecontrol systems.

See also: --qp, --bitrate

[edit] rc-lookahead

Default: 40

Sets the number of frames to use for mb-tree ratecontrol and vbv-lookahead. The maximum allowed value is 250.

For the mb-tree portion of this, increasing the frame count generates better results but is also slower. The maximum buffer value used by mb-tree is the MIN( rc-lookahead, --keyint )

For the vbv-lookahead portion of this, increasing the frame count generates better stability and accuracy when using vbv. The maximum value used by vbv-lookahead is: 

MIN(rc-lookahead, MAX(--keyint, MAX(--vbv-maxrate, --bitrate) / --vbv-bufsize * --fps))

See Also: --no-mbtree, --vbv-bufsize, --vbv-maxrate

[edit] vbv-maxrate

Default: 0

Sets the maximum rate the VBV buffer should be assumed to refill at.

VBV reduces quality, so you should only use this if you're encoding for a playback scenario that requires it.

See also: --vbv-bufsize, --vbv-init, VBV Encoding Suggestions

[edit] vbv-bufsize

Default: 0

Sets the size of the VBV buffer in kilobits.

VBV reduces quality, so you should only use this if you're encoding for a playback scenario that requires it.

See also: --vbv-maxsize, --vbv-init, VBV Encoding Suggestions

[edit] vbv-init

Default: 0.9

Sets how full the VBV Buffer must be before playback starts.

If it is less than 1, the the initial fill is: vbv-init * vbv-bufsize. Otherwise it is interpreted as the initial fill in kbits.

See also: --vbv-maxsize, --vbv-bufsize, VBV Encoding Suggestions

[edit] crf-max

Default: Not set

A similar setting to --qpmax except instead of specifying a maximum quantizer you're specifying a maximum ratefactor. This option only works when you are using CRF and have VBV enabled. It prevents x264 from reducing the ratefactor (aka "quality") below the given value even when doing so would violate VBV constraints. This setting is mostly applicable to custom streaming servers. More information can be found in the initial commit message.

See Also: --crf, --vbv-maxrate, --vbv-bufsize

[edit] qpmin

Default: 0

Defines the minimum quantizer that x264 will ever use. The lower the quantizer, the closer the output is to the input. At some point, the output of x264 will look the same as the input, even though it is not exactly the same. Usually there is no reason to allow x264 to spend more bits than this on any particular macroblock.

With adaptive quantization enabled (the default), raising qpmin is discouraged because this could reduce the quality of flat background areas of the frame.

See also: --qpmax, --ipratio

[edit] qpmax

Default: 51

The opposite of qpmin, above. Defines the maximum quantizer that x264 can use. The default of 51 is the highest quantizer available for use in the H.264 spec, and is extremely low quality. This default effectively disables qpmax. You may want to set this lower (values in the 30-40 range are generally as low as you'd go) if you want to cap the minimum quality x264 can output, but adjusting it is generally not recommended.

See also: --qpmin, --pbratio, --crf-max

[edit] qpstep

Default: 4

Sets the maximum change in quantizer between two frames.

[edit] ratetol

Default: 1.0

This is a dual purpose parameter:

  • In 1-pass bitrate encodes, this settings controls the percentage that x264 can miss the target average bitrate by. You can set this to 'inf' to disable this overflow detection completely. The lowest you can set this is to 0.01. The higher you set this to the better x264 can react to complex scenes near the end of the movie. The unit of measure for this purpose is percent (eg, 1.0 = 1% bitrate deviation allowed).
Many movies (any action movie, for instance) are most complex at the climatic finale. As a 1pass encode doesn't know this, the number of bits required for the end is usually underestimated. A ratetol of inf can mitigate this by allowing the encode to function more like a --crf encode, but the filesize will blow out.
  • When VBV is activated (ie, you're specified --vbv-* options), this setting also affects VBV aggressiveness. Setting this higher allows VBV to fluctuate more at the risk of possibly violating the VBV settings. For this purpose, the unit of measure is arbitrary.

[edit] ipratio

Default: 1.40

Modifies the target average increase in quantizer for I-frames as compared to P-frames. Higher values increase the quality of I-frames generated.

See also: --pbratio

[edit] pbratio

Default: 1.30

Modifies the target average decrease in quantizer for B-frames as compared to P-frames. Higher values decrease the quality of B-frames generated. Not used with mbtree (enabled by default), which calculates the optimum value automatically.

See also: --ipratio

[edit] chroma-qp-offset

Default: 0

Add an offset to the quantizer of chroma planes when encoding. The offset can be negative.

x264 automatically lowers this when using psy-rd and/or psy-trellis to compensate for it increasing the quality of luma, and subsequently decreasing the quality of chroma. The default values for these settings cause 2 to be subtracted from the chroma-qp-offset.

Note: x264 only encodes the luma and chroma planes at the same quantizer up to quantizer 29. After this, chroma is progressively quantized by a lower amount than luma until you end with luma at q51 and chroma at q39. This behavior is required by the H.264 standard.

[edit] aq-mode

Adaptive Quantization Mode

Default: 1

Without AQ, x264 tends to underallocate bits to less-detailed sections. AQ is used to better distribute the available bits between all macroblocks in the video. This setting changes what scope AQ re-arranges bits in:

  • 0: Do not use AQ at all.
  • 1: Allow AQ to redistribute bits across the whole video and within frames.
  • 2: Auto-variance AQ (experimental) which attempts to adapt strength per-frame.

See also: --aq-strength

[edit] aq-strength

Adaptive Quantization Strength

Default: 1.0

Sets the strength of AQ bias towards low detail ('flat') macroblocks. Negative values are not allowed. Values more than +/-100% from 1.0 are probably a bad idea.

See also: --aq-mode

[edit] pass

Default: Not Set

This is an important setting for 2pass encoding. It controls what x264 will do with the --stats file. It has three settings:

  • 1: Create a new stats file. Use this on the first pass.
  • 2: Read the stats file. Use this on the final pass.
  • 3: Read the stats file, and update it as well.

The stats file contains information about every input frame, which can be input to x264 in order to improve the output. The idea is you run a first pass to generate the stats file, and the second pass will create an optimized encode of the video. The improvement is mostly gained from better ratecontrol.

See also: --stats, --bitrate, --slow-firstpass X264_statsfile

[edit] stats

Default: 'x264_2pass.log'

Set the location that x264 will read/write the --pass X264_statsfile to/from.

See also: --pass

[edit] no-mbtree

Default: Not Set

Disable macroblock tree ratecontrol. Using macroblock tree ratecontrol overall improves the compression by keeping track of temporal propagation across frames and weighting accordingly. Requires a new large statsfile in addition to the already existing for multipass encodes.

Recommendation: Default

See Also: --rc-lookahead

[edit] qcomp

Default: 0.60

Quantizer curve compression factor. 0.0 => Constant Bitrate, 1.0 => Constant Quantizer.

When used with mbtree, it affects the strength of mbtree. (Higher qcomp = weaker mbtree).

Recommendation: Default

See Also: --cplxblur, --qblur

[edit] cplxblur

Default: 20

Apply a gaussian blur with the given radius to the quantizer curve. This means that the quantizer assigned to each frame is blurred temporally with its neighbours to limit quantizer fluctuations.

See also: --qcomp, --qblur

[edit] qblur

Default: 0.5

Apply a gaussian blur with the given radius to the quantizer curve, after curve compression. Not a very important setting.

See also: --qcomp, --cplxblur

[edit] zones

Default: Not Set

Tweak settings for specific sections of the video. You can modify most x264 options per-zone.

  • A single zone takes the form of <start frame>,<end frame>,<options>
  • Multiple zones are separated from each other with a '/'


Options:

These two options are special. You can only set one per zone, and if you set one, it must be the first option listed for the zone:

  • b=<float> applies a bitrate multiplier on the zone. Useful for extra tweaking of high- and low-action scenes.
  • q=<int> applies a constant quantizer on the zone. Useful for applying to a range of frames.

The other available options are as follows:

  • ref=<integer>
  • b-bias=<integer>
  • scenecut=<integer>
  • no-deblock
  • deblock=<integer>:<integer>
  • deadzone-intra=<integer>
  • deadzone-inter=<integer>
  • direct=<string>
  • merange=<integer>
  • nr=<integer>
  • subme=<integer>
  • trellis=<integer>
  • (no-)chroma-me
  • (no-)dct-decimate
  • (no-)fast-pskip
  • (no-)mixed-refs
  • psy-rd=<float>:<float>
  • me=<string>
  • no-8x8dct
  • b-pyramid=<string>

Limitations:

  • The number of reference frames for a zone can never exceed what was specified with --ref
  • Scenecut can not be turned on and off; only varied if originally active (>0)
  • Merange can not exceed what was originally specified if --me esa/tesa
  • Subme can't be changed if the original commandline specified it as 0.
  • You can't set me to esa or tesa if --me was originally specified as dia, hex, or umh

Example: 0,1000,b=2/1001,2000,q=20,me=3,b-bias=-1000

Recommendation: Default

[edit] qpfile

Manual override to standard ratecontrol. Specify a file that gives the quantizer and frametype for specified frames. The format is 'framenum frametype quantizer'. For example: 

0 I 18 < IDR (key) I-frame
1 P 18 < P-frame
2 B 18 < Referenced B-frame
3 i 18 < Non-IDR (non-key) I-frame
4 b 18 < Non-referenced B-frame
5 K 18 < Keyframe*
  • You don't need to specify every frame
  • Using -1 as the desired quantizer allows x264 to choose the optimal quantizer value, useful when only needing to set the frametype
  • Having a large number of specified frame types and quantizers while still letting x264 choose intermittently decreases x264's performance
  • 'Keyframe' is a generic keyframe/seekpoint type that equates to a IDR I-Frame if --open-gop is none, otherwise it equates to a Non-IDR I-Frame flagged with the Recovery Point SEI

[edit] Analysis

[edit] partitions

Default: 'p8x8,b8x8,i8x8,i4x4'

H.264 video is split up into 16x16 macroblocks during compression. These blocks can be further split up into smaller partitions, which is what this option controls.

With this option, you enable individual partitions. Partitions are enabled per-frametype (i.e., I, P, B). The available partitions are p8x8, p4x4, b8x8, i8x8, and i4x4.

  • I: i8x8, i4x4
  • P: p8x8 (also enables p16x8/p8x16), p4x4 (also enables p8x4/p4x8)
  • B: b8x8 (also enables b16x8/b8x16)

You can also set 'none' or 'all'.

p4x4 is generally not very useful and has an extremely high ratio of speed cost to resulting quality gain.

See also: --no-8x8dct

[edit] direct

Default: 'spatial'

Set prediction mode for 'direct' motion vectors. There are two modes available: spatial and temporal. You can also select none to disable direct MVs, and auto to allow x264 to swap between them as it sees fit. If you set auto, x264 outputs information on the usage at the end of the encode. 'auto' works best in a 2pass encode, but will work in single-pass encodes too. In first-pass auto mode, x264 keeps a running average of how well each method has so far performed, and picks the next prediction mode from that. Note that you should only enable auto on the second pass if it was enabled on the first pass; if it wasn't, the second pass will default to temporal. Direct none wastes bits and is strongly discouraged.

Recommendation: 'auto'

[edit] no-weightb

Default: Not Set

H.264 allows you to 'weight' references in B-frames, which allows you to change how much each reference affects the predicted picture. This disables that feature.

Recommendation: Default

[edit] weightp

Default: 2

Enables use of explicit weighted prediction to improve compression in P-frames. Higher modes are slower.

NOTE: When encoding for Adobe Flash set this to 0 - its decoder generates artifacts otherwise. Flash 10.1 fixes this bug.

Modes:

  • 0: Disabled
  • 1: Blind offset (a constant offset of -1)
  • 2: Smart analysis with duplicates, designed specifically to improve compression in fades.

[edit] me

Default: 'hex'

Set the full-pixel motion estimation method. There are five choices:

  • dia (diamond) is the simplest search, consisting of starting at the best predictor, checking the motion vectors at one pixel upwards, left, down, and to the right, picking the best, and repeating the process until it no longer finds any better motion vector.
  • hex (hexagon) consists of a similar strategy, except it uses a range-2 search of 6 surrounding points, thus the name. It is considerably more efficient than dia and hardly any slower, and therefore makes a good choice for general-use encoding.
  • umh (uneven multi-hex) is considerably slower than hex, but searches a complex multi-hexagon pattern in order to avoid missing harder-to-find motion vectors. Unlike hex and dia, the merange parameter directly controls umh's search radius, allowing one to increase or decrease the size of the wide search.
  • esa (exhaustive) is a highly optimized intelligent search of the entire motion search space within merange of the best predictor. It is mathematically equivalent to the bruteforce method of searching every single motion vector in that area, though faster. However, it is still considerably slower than UMH, with not too much benefit, so is not particularly useful for everyday encoding.
  • tesa (transformed exhaustive) is an algorithm which attempts to approximate the effect of running a Hadamard transform comparison at each motion vector; like exhaustive, but a little bit better and a little bit slower.

See also: --merange

[edit] merange

Default: 16

merange controls the max range of the motion search in pixels. For hex and dia, the range is clamped to 4-16, with a default of 16. For umh and esa, it can be increased beyond the default 16 to allow for a wider-range motion search, which is useful on HD footage and for high-motion footage. Note that for umh, esa, and tesa, increasing merange will significantly slow down encoding.

See also: --me

[edit] mvrange

Default: -1 (auto)

Set the maximum (vertical) range of any one motion vector in pixels. The default value is level-dependent:

  • Level 1/1b: 64
  • Level 1.1-2.0: 128
  • Level 2.1-3.0: 256
  • Level 3.1+: 512

Note: if you want to manually override the mvrange, subtract 0.25 from the above values when setting (eg --mvrange 127.75).

Recommendation: Default

[edit] mvrange-thread

Default: -1 (auto)

Set the minimum motion vector buffer between threads. Don't touch it.

Recommendation: Default

[edit] subme

Default: 7

Set the subpixel estimation complexity. Higher numbers are better. Levels 1-5 simply control the subpixel refinement strength. Level 6 enables RDO for mode decision, and level 8 enables RDO for motion vectors and intra prediction modes. RDO levels are significantly slower than the previous levels.

Using a value less than 2 will enable a faster, and lower quality lookahead mode, as well as cause poorer --scenecut decisions to be made, and thus it is not recommended.

Possible Values:

0. fullpel only
1. QPel SAD 1 iteration
2. QPel SATD 2 iterations
3. HPel on MB then QPel
4. Always QPel
5. Multi QPel + bi-directional motion estimation
6. RD on I/P frames
7. RD on all frames
8. RD refinement on I/P frames
9. RD refinement on all frames
10. QP-RD (requires --trellis=2, --aq-mode > 0)

Recommendation: Default, or higher, unless speed is very important.

[edit] subq

Alias of --subme

[edit] psy-rd

Default: 1.0:0.0

The first number is the strength of Psy-RDO to use (requires subme>=6 to activate). The second number is the strength of Psy-Trellis (requires trellis>=1 to activate). Note that Trellis is still considered 'experimental', and almost certainly is a Bad Thing for at least cartoons.

See this thread on doom9 for an explanation of psy-rd.

[edit] no-psy

Default: Not Set

Disables all visual optimizations that reduce PSNR or SSIM. This also disables some internal psy optimizations that aren't settable via x264's command line arguments.

Recommendation: Default

[edit] no-mixed-refs

Default: Not Set

Mixed refs will select refs on a per-8x8 partition, rather than per-macroblock basis. This improves quality when using multiple reference frames, albeit at some speed cost. Setting this option will disable it.

Recommendation: Default

See also: --ref

[edit] no-chroma-me

Default: Not Set

Normally, motion estimation works off both the luma and chroma planes. This disables chroma motion estimation for a small speed boost.

Recommendation: Default

[edit] no-8x8dct

Default: Not Set

Adaptive 8x8 DCT enables the intelligent adaptive use of 8x8 transforms in I-frames. This disables the feature.

Recommendation: Default

[edit] trellis

Default: 1

Performs Trellis quantization to increase efficiency.

0. Disabled
1. Enabled only on the final encode of a macroblock
2. Enabled on all mode decisions

On Macroblock provides a good compromise between speed and efficiency. On all decisions reduces speed further.

See: Trellis Quantization

Recommendation: Default

Note: Requires --cabac

[edit] no-fast-pskip

Default: Not Set

Disables early skip detection on P-frames. Very slightly increases quality at a large speed cost.

Recommendation: Default

[edit] no-dct-decimate

Default: Not Set

DCT Decimation will drop DCT blocks it deems "unnecessary". This will improve coding efficiency, with a usually negligible loss in quality. Setting this option will disable it.

Recommendation: Default

[edit] nr

Default: Not Set

Performs fast noise reduction. Estimates film noise based on this value and attempts to remove it by dropping small details before quantization. This may not match the quality of a good external noise reduction filter, but it performs very fast.

Recommendation: Default or (100 to 1000 for denoising)

[edit] deadzone-inter/intra

Default: Not Set

Set the size of the inter/intra luma quantization deadzone. Deadzones should be in the range of 0 to 32. The deadzone value sets the level of fine detail that x264 will arbitrarily drop without attempting to preserve. Very fine detail is both hard to see and expensive to encode, dropping this detail without attempting to preserve it stops wasting bits on such a low-return section of the video. Deadzone is incompatible with Trellis.

Recommendation: Default

[edit] cqm

Default: Flat (Not Set)

Sets all custom quantization matrices to those of a built-in preset. The built-in presets are flat or JVT.

Recommendation: Default

See also: --cqmfile

[edit] cqmfile

Default: Not Set

Sets custom quantization matrices from a specified JM-compatible file. Overrides any other --cqm* options.

Recommendation: Default

See also: --cqm

[edit] cqm4* / cqm8*

Default: Not Set

  • --cqm4: Set all 4x4 quant matrices. Takes a comma-separated list of 16 integers.
  • --cqm8: Set all 8x8 quant matrices. Takes a comma-separated list of 64 integers.
  • --cqm4i, --cqm4p, --cqm8i, --cqm8p: Set both luma and chroma quant matrices
  • --cqm4iy, --cqm4ic, --cqm4py, --cqm4pc: Set individual quant matrices. Same switches exist for cqm8.

Recommendation: Default

[edit] Video Usability Info

These options set a flag in the output stream that can be read by the decoding application and possibly acted on. It's worth noting that most of these options in most scenarios are pointless, and are usually ignored by software decoders.

[edit] overscan

Default: undef

How to handle overscan. Overscan is used here in the sense of a device only displaying part of an image.

Possible Values:

  • undef - Undefined.
  • show - Indicate to show the entire image. Theoretically must be respected if set.
  • crop - Indicate that the image is suitable for playback on devices with overscan. Not necessarily respected.

Recommendation: Crop before encoding and use show if your device supports it, otherwise ignore.

[edit] videoformat

Default: undef

Indicates what the video was before encoding/digitizing.

Possible Values:

  • component
  • pal
  • ntsc
  • secam
  • mac
  • undef

Recommendation: Whatever your source video was, or undefined.

[edit] fullrange

Default: off

Indicates whether to use the full range of luma and chroma levels. If set to off, the limited ranges will be used.

See this page for a simple description.

Recommendation: If your source is digitized from analog video, then set this to off. Otherwise, set it to on.

[edit] colorprim

Default: undef

Set what color primaries for converting to RGB.

Possible Values:

  • undef
  • bt709
  • bt470m
  • bt470bg
  • smpte170m
  • smpte240m
  • film

See: RGB and YCrCb

Recommendation: Default, unless you know what your source uses.

[edit] transfer

Default: undef

Set the opto-electronic transfer characteristics to use. (Sets the gamma curve to use for correction.)

Possible Values

  • undef
  • bt709
  • bt470m
  • bt470bg
  • linear
  • log100
  • log316
  • smpte170m
  • smpte240m

See: Gamma Correction

Recommendation: Default, unless you know what your source uses.

[edit] colormatrix

Default: undef

Set the matrix coefficients used in deriving the luma and chroma from the RGB primaries.

Possible Values

  • undef
  • bt709
  • fcc
  • bt470bg
  • smpte170m
  • smpte240m
  • GBR
  • YCgCo

See: YCbCr

Recommendation: Whatever your sources uses, or default.

[edit] chromaloc

Default: 0

Sets the chroma sample location. (as defined in Annex E of the ITU-T Specification).

Values range from 0 to 5.

See x264's vui.txt

Recommendation:

  • If you transcode from MPEG1 with proper subsampled 4:2:0, and don't do any color space conversion, you should set this option to 1.
  • If you transcode from MPEG2 with proper subsampled 4:2:0, and don't do any color space conversion, you should set this option to 0.
  • If you transcode from MPEG4 with proper subsampled 4:2:0, and don't do any color space conversion, you should set this option to 0.
  • Otherwise, default.

[edit] nal-hrd

Default: None

Signal HRD information. Required for Blu-ray streams, television broadcast and a few other specialist areas. Acceptable values are:

  • none Specify no HRD information
  • vbr Specify HRD information
  • cbr Specify HRD information and pack the bitstream to the bitrate specified by bitrate. Requires bitrate mode ratecontrol.

Recommendation: none, unless you need to signal this information.

See also: --vbv-bufsize, --vbv-maxrate, --aud

[edit] pic-struct

Default: Not Set

Force sending pic_struct in Picture Timing SEI.

Implied when you use --pulldown or --tff/--bff.

Recommendation: Default

[edit] Input/Output

[edit] output

Default: Not Set.

Specifies output filename. The extension you specify determines the output format of your video. If the extension is not recognised the default output format is the raw video stream (generally stored with the .264 extension).

The special location NUL (Windows) or /dev/null (Unix) specifies the output should be discarded. This is particularly useful when using pass 1, as the only output you care about is that from stats.

[edit] muxer

Default: auto

Specifies what format to write to.

Possible Values

  • auto
  • raw
  • mkv
  • flv
  • mp4

The 'auto' option will pick a muxer based on the supplied output filename.

See also: --output

Recommendation: Default

[edit] demuxer

Default: Automatically detected.

Sets what demuxer and decoder x264 uses for parsing the input video.

Possible Values

  • auto
  • raw
  • y4m
  • avs
  • lavf
  • ffms

If the input file has an extension of raw, y4m or avs, x264 will use the relevant demuxer to read the file. Standard input uses the raw demuxer. Otherwise, x264 will attempt to open the file with ffms, then lavf, and then fail.

The 'lavf' and 'ffms' options require x264 to be compiled with the respective libraries. If either is used, x264 will carry over the timecodes from the input file, provided you don't output to raw. This effectively makes x264 VFR-aware. The other options can have a constant framerate specified with --fps or a variable framerate with --tcfile-in.

See also: --input, --muxer

Recommendation: Default.

[edit] input-csp

Default: i420

Tell x264 what colourspace your raw video input is with this switch. Supported colourspaces are listed in x264 --fullhelp.

Note that while RGB colourspaces are listed, the video is converted to YUV using the bt601 (ie, "SD") matrix before encoding.


See Also: --input-res, --fps

[edit] input-res

Specify the input resolution of raw video input. Use the syntax --input-res 720x576.

See Also: --input-csp, --fps

[edit] index

Default: Not Set

An optional setting that only takes effect when using the ffms --demuxer. Specifies a file for ffms to write out indexing data for the input file to, which can be referred to in a future encode to remove the need to reindex the video. Generally not needed -- indexing is not a slow process relative to video encoding.

See also: --demuxer, FFMS2 API Documentation

Recommendation: Default, unless you want to save a minute amount of time indexing.

[edit] sar

Default: Not Set

Specifies the input video's Sample Aspect Ratio (SAR) to be used by the encoder in width:height. This in conjunction with frame dimensions can be used to encode an anamorphic output by determining the Display Aspect Ratio (DAR) via the formula: DAR = SAR x width/height

See Main Article here Recommendation: You might need to set this if you're using the resize filter and encoding with anamorphic input.

[edit] fps

Default: autodetected

Specifies video framerate as either a float (29.970) a rational (30000/1001), or an integer (2997/100) value. x264 detects and uses the framerate from the input stream header when available (y4m, avs, ffms and lavf demuxer), otherwise uses 25. Setting this implies force-cfr.

If you are using raw YUV input and --bitrate-based ratecontrol, you need to specify the correct framerate using this parameter or --tcfile-in. x264 won't hit your target bitrate otherwise.

[edit] seek

Default: Not Set

Specifies the first frame to encode, allowing the encode to begin at any point in the source.

Recommendation: Default

[edit] frames

Default: Not Set

Specifies the maximum number of frames to encode, allowing the encode to finish at any point before the end of the source.

Recommendation: Default

[edit] level

Default: -1 (auto)

Sets the level flag in the output bitstream (as defined by Annex A of the H.264 standard). Permissible levels are:

1 1.1 1.2 1.3 2 2.1 2.2 3 3.1 3.2 4 4.1 4.2 5 5.1

(Level 1b is not a supported value.)

If you do not specify --level on the commandline, x264 will attempt to autodetect the level. This detection is not perfect and may underestimate the level if you are not using VBV. x264 will also automatically limit the DPB size (see --ref) to remain in compliance with the level you select (unless you also manually specify --ref).

What Level Do I Pick?

Level 4.1 is often considered the highest level you can rely on desktop consumer hardware to support. Blu-ray Discs only support level 4.1, and many non-mobile devices like the Xbox 360 specify level 4.1 as the highest they officially support. Mobile devices like the iPhone/Android are a totally different story.

Wikipedia has a nice chart detailing the restrictions for each level, if you want to read it.

Recommendation: Default, unless you are aiming for a specific device.

[edit] verbose

Default: Not Set

Displays statistics for each encoded frame.

Recommendation: Default

[edit] no-progress

Default: Not Set

Disables the progress indicator while encoding.

Recommendation: Default

[edit] quiet

Default: Not Set

Enables Quiet Mode, silencing status messages from x264.

Recommendation: Default

[edit] log-level

Default: info

Manually specify the logging level of both x264cli and libx264.

Possible Values

  • none
  • error
  • warning
  • info
  • debug

Recommendation: Default

[edit] psnr

Default: Not Set

Enables PSNR calculations that are reported on completion at the cost of a small decrease in speed.

Recommendation: Default

[edit] ssim

Default: Not Set

Enables SSIM calculations that are reported on completion at the cost of a small decrease in speed.

Recommendation: Default

[edit] threads

Default: auto (frame based threads: 1.5 * logical processors, rounded down; slice based threads: 1 * logical processors)

Enables parallel encoding by using more than 1 thread to increase speed on multi-core systems. The quality loss from multiple threads is mostly negligible unless using very high numbers of threads (say, above 16). The speed gain should be slightly less than linear until you start using more than 1 thread per 40px of vertical video, at which point the gain from additional threads sharply decreases.

x264 currently has an internal limit on the number of threads set at 128, realistically you should never set it this high.

Recommendation: Default

See also: thread-input, sliced-threads

[edit] sliced-threads

Default: off

Enables slice-based threading. This threading method produces lower quality results than the default method both compression and efficiency-wise, but adds no encoding latency.

The maximum number of sliced threads is MIN( (height+15)/16 / 4, 128 )

Recommendation: Default (off), unless you are doing some sort of realtime streaming or low latency is important.

[edit] thread-input

Default: Set if threads > 1.

Decodes the input video in a separate thread to the encoding process.

Recommendation: Default.

[edit] sync-lookahead

Default: auto (bframes+1)

Sets the number of frames to be used as a buffer for threaded lookahead. Maximum Value is 250. Automatically disabled during the 2nd or greater pass or when using sliced threads.

Setting this to 0 disables threaded lookahead, which allows lower latency at the cost of reduced performance.

Recommendation: Default

[edit] non-deterministic

Default: Not Set

Slightly improve quality when encoding with --threads > 1, at the cost of non-deterministic output encodes. This enables multi-threaded mv and uses the entire lookahead buffer in slicetype decisions when slicetype is threaded -- rather than just the minimum amount known to be available.

Not for general use.

Recommendation: Default

See also: threads

[edit] asm

Default: auto

Override automatic CPU detection. Useful for debugging or troubleshooting.

Recommendation: Default

[edit] no-asm

Default: Not Set

Disables all CPU optimisations. Useful for debugging or troubleshooting.

Recommendation: Default

[edit] visualize

Default: Not Set

Enables Macroblock Type visualizations over the encoded video. Useful for frame by frame debugging or analysis.

Requires compile time support, and an X11 windowing system.

Recommendation: Default

[edit] dump-yuv

Default: Not Set

Dumps the reconstructed YUV frames to the specified file. Useful mostly for debugging. Not for general use.

Recommendation: Default

[edit] sps-id

Default: Not Set

Set SPS (sequence parameter set) and PPS (picture parameter set) id numbers. Not for general use.

Recommendation: Default

[edit] aud

Default: Not Set

Use access unit delimiters.

Recommendation: Default, unless encoding for Blu-ray, in which case set this option.

[edit] force-cfr

Default: Not Set

If using ffms2 or lavf demuxers, timecodes are copied from the input file, provided the output file is not raw. This option disables this, and forces x264 to generate its own. When using this you probably also want to set --fps.

Recommendation: Default

[edit] tcfile-in

Specify a timecodes file which should be used to interpret the framerate of the input video. The timecode file can be in either v1 or v2 format which are described in the mkvmerge documentation.

See Also: tcfile-out, force-cfr, fps

[edit] tcfile-out

Output a timecode file (v2 format) based on the input timestamps. For use when you're encoding a VFR input video and want to dump the timecodes. For details on the file format, see tcfile-in.

[edit] timebase

Default: Not Set

Allows you to set a custom timebase.

The numerator is 'seconds', and the denominator is 'ticks'. Means how many seconds one tick takes.

  • If passed a rational, it will set the numerator and denominator accordingly.
  • If passed an integer, and an input timecodes file is set via tcfile-in, it will use the value passed as the numerator, and generate the denominator accordingly.
  • If passed an integer, and no input timecodes file is set, it will use the value passed as the denominator and generate the 'ticks per frame' from the input video.

Not compatible with --force-cfr mode.

Recommendation: Default

[edit] dts-compress

Default: Not Set

A hack, implemented for FLV and MP4 containers only, that can work around buggy decoders that expect all DTSs to be positive. As per the commit message for this change, use at your own risk.

Note: DTS refers to Decode TimeStamp. Each frame is allocated a DTS that reflects its position in the 'coded order' of the stream, as opposed to the 'display order' of the stream, which is specified by the Presentation TimeStamp. The order frames are stored in the stream is different to the order they are displayed due to compression techniques such as B-frames, which result in frames requiring data from frames that will be displayed after.

[edit] Filtering

[edit] video-filter

The x264 filtering system is used to process the input video before encoding. Multiple filters can also be applied sequentially to the input video.

The basic syntax to apply a filter is this: 

--video-filter <filter>

You can apply multiple filters sequentially by delimiting them with a /: 

--video-filter <filter1>/<filter2>

You can 'chain' as many filter operations as you like together.

The available filters are:

[edit] crop

Syntax: crop:left,top,right,bottom

Remove pixels from the edges of the frame.

  • You must crop in multiples of 2 when using yv12, i420, or nv12 format video input.
  • You must crop width in multiples of 2 when using i422 format video input.

[edit] resize

Syntax: resize:[width,height][,sar][,fittobox][,csp][,method]

Resizes frames and/or converts frames between different colorspaces. Requires x264 to be compiled with libswscale.

There are a few possible modes on how to resize frames:

  • Resolution only: Resizes frames to the specified resolution and changes the SAR to avoid stretching.
  • SAR only: Sets the SAR and resizes frames to a new resolution to avoid stretching.
  • Resolution + SAR: Resizes frames to the specified resolution and sets SAR to the specified value, potentially allowing stretching.
  • Fittobox: Resizes frames based on the specified constraint mode, adapting the resolution to have a resultant SAR of 1:1
  • width: Resize frames to fit within the specified width constraint
  • height: Resize frames to fit within the specified height constraint
  • both: Resize frames to fit within the specified box constraint
  • Fittobox + SAR: Same as regular Fittobox mode except the resultant frames have the specified SAR, shrinking the video to where the anamorphic video fits within the specified constraint.

Options that are independent of the resizing mode are

  • csp: simultaneously converting frames to the specified colorspace. The valid list of colorspaces are displayed in x264 --fullhelp
  • method (Default bicubic): when resizing frames, use the specified resizer method
fastbilinear, bilinear, bicubic, experimental, point, area, bicublin, gauss, sinc, lanczos, spline

Example: 

resize:width=1280,height=720,method=spline

[edit] select_every

Syntax: select_every:step,offset1[,offset2,...]

"Select" only a subset of input frames to encode, dropping the rest. Every step frames, take only the frames specified with an offset. For example: To encode every second frame: 

select_every:2,1

To drop every third frame: 

select_every:3,0,1

You can see more examples at the Avisynth wiki (which has an identical filter).

[edit] See Also

Retrieved from "http://mewiki.project357.com/wiki/X264_Settings"
Personal tools