9.7 KiB
Aomenc
GIT: AOM
Table of Contents
Important command line options
| Command Line | Description |
|---|---|
| --help | Show usage options and exit |
| --end-usage=arg | Rate control mode (vbr, cbr(default), cq, q). VBR and CBR are self explanatory. CQ is constant quantizer with a bitrate ceiling(Constrained Quality), Q is for Quality. |
| --cq-level=arg | Constant/Constrained Quality level, used in Q/CQ modes. |
| --cpu-used=arg | CPU Used (0..6) Good mode, (5..9) realtime mode 1(default). Default is good mode(CPU-0 to CPU-6): unless you specify the realtime flag(--rt), every speed set above 6 will default back to 6. Lower numbers are slower. |
| --target-bitrate=arg | Bitrate (kbps) |
| --bit-depth=arg | Bit depth (8, 10, 12). Default is the bit-depth recognized by aomenc from the source file. 12-bit is not recommended for end compression, as it is not supported in the main AV1 HW decoding profile. |
| --tile-columns=arg | Number of tile columns to use, log2 (number to power of 2). With --tile-columns=2, it'll do 2²= 4 tile columns. |
| --tile-rows=arg | Number of tile rows to use, log2 (number to power of 2). With --tile-rows=1, it'll do 2¹= 2 tile rows. |
| --threads=arg | Allowed number of threads to use. |
| --lag-in-frames=arg | Number of lagged frames used by the encoder for lookahead and alternate reference frame placement(default 19, max 35). |
| --enable-cdef=arg | Enable the constrained directional enhancement filter (0: false, 1: true (default)). CDEF is a very powerful filter used to clean up ringing, haloing and ringing artifacts. It works very well in that regard. |
| --aq-mode=arg | Adaptive quantization mode(0: default. 1: Variance. 2: Complexity. 3: Cyclic Refresh |
| --tune-content=arg | Tune content type (default,screen). |
| --enable-fwd-kf=arg | Enable forward reference keyframes(default=0). |
| --kf-min-dist=arg | Minimum keyframe interval in frames(default=12). |
| --kf-max-dist=arg | Maximum keyframe interval in frames(default=9999, or adaptive keyframe placement only). |
| --enable-keyframe-filtering=arg | Apply temporal filtering on key frame(0: no filter, 1: filter without overlay (default), 2: filter with overlay - experimental, may break random access in players.). It is recommended to leave it at default unless you really know what you're doing. |
| --arnr-maxframes=arg | Maximum number of alternate reference noise reduced frames used by the encoder(default=7). |
| --arnr-strength=arg | ARNR frames filtering strength(default=5). |
| --enable-qm=arg | Enable quantisation matrices (0: false (default), 1: true). |
| --quant-b-adapt=arg | Use adaptive quantize_b(default=0). |
| --mv-cost-upd-freq=arg | Update freq for mv costs(motion vector estimation cost calculation) 0: SB(SuperBlock), 1: SB Row per Tile, 2: Tile, 3: Off. |
| --enable-chroma-deltaq=arg | Enable chroma delta quant (0: false (default), 1: true). May be broken below --cq-level=15. |
| --color-primaries=arg | Color primaries (CICP) of input content: bt709, unspecified, bt601, bt470m, bt470bg, smpte240, film, bt2020, xyz, smpte431, smpte432, ebu3213. |
| --transfer-characteristics=arg | Transfer characteristics (CICP) of input content(unspecified, bt709, bt470m, bt470bg, bt601, smpte240, lin, log100, log100sq10, iec61966, bt1361, srgb, bt2020-10bit, bt2020-12bit, smpte2084, hlg, smpte428. |
| --matrix-coefficients=arg | Matrix coefficients (CICP) of input content: identity, bt709, unspecified, fcc73, bt470bg, bt601, smpte240, ycgco, bt2020ncl, bt2020cl, smpte2085, chromncl, chromcl, ictcp. |
Example settings and explanation
Constant quality:
--end-usage=q --cq-level=30 --cpu-used=4 --threads=16
It is recommended to set it the rate control --end-usage=q to get the highest quality rate control method possible. Only use CQ if you are planning to stream with a maximum bitrate, and CBR for livestreaming.
It is recommended to the --cq-level in range 20-40 depending on your source.
Target Bitrate:
--end-usage=vbr --target-bitrate=1000 --cpu-used=4 --threads=16
To get good efficiency with VBR, it is strongly recommended to use aomenc in 2-pass mode(which is the default in av1an).
Tiles(tile columns and rows)
... --tile-columns=2 --tile-rows=1 ...
For highest efficiency while keeping good threading, it is recommended to set it to --tile-columns=2 and --tile-rows=1 at 1080p. For higher resolution/higher framerate encoding, set it to --tile-columns=3 and --tile-rows=2.
CPU preset:
... --cpu-used=6 ...
--cpu-used=6 is recommended if you want good compression efficiency and fast encoding, and --cpu-used=4 if you want to crank up the efficiency even further. It is not recommended to use slower presets as they are much slower without much gain.
Bit-depth:
... --bit-depth=10 ...
It is recommended to set it to 10-bit even for 8-bit content for higher efficiency and less banding due to no precision losses going from an 8-bit YUV source to 8-bit YUV, and allows for higher calculation precision. However, as of January 23rd 2021, it is not very easy to decode 10-bit AV1 video on x86_64 CPUs due to a lack of assembly optimizations. For now, use 10-bit encoding for 1080p30 content. Otherwise, use 8-bit with --bit-depth=8.
Threading and tile threading
... --threads=8 --tile-columns=1 --tile-rows=0 ... or ... --threads=16 --tile-columns=2 --tile-rows=1 ... for single instance encoding.
If you have say an 8C/16T CPU, it is recommended to set to #cores/2 you have for chunked encoding. If you have a higher thread count CPU, limiting it to 4 threads is also a good option to prevent thread oversaturation. Give it how many threads you have for single worker encoding.
Lag-in-frames
... --lag-in-frames=35 ...
More is better, but it makes the encoder slower, up to a limit of 35 (default is 19).
CDEF usage
... --enable-cdef=0 ... if you want to encode >1080p30 10-bit as of January 2021.
For video game content and content in 8-bit, I always recommend keeping it on, as it makes the image much cleaner overall by preventing aliasing from ruining the imag, and gives a nice boost to subjective quality. However, for 10-bit AV1 on x86_64 CPUs, there is currently a big decoding performance penalty, so if you have >1080p30 footage or want maximum decode performance NOW, you can set --enable-cdef=0.
Adaptive quantization mode
... --aq-mode=2 ...
Adaptive quantization mode (0: Default. 1: Variance, in which the encoder lowers the quantizer value for flat low complexity blocks, like clouds or dark scenes. Not recommended unless you know what you are doing. 2: Complexity, in which the encoder lowers the quantizer for complex parts of the image. 3: Cyclic refresh. Ups the quantizer value as much as possible for static flat looking blocks, only doing block refreshes in cycles if needed. Not recommend unless when used for live-streaming or conferencing.)
Content tune
... --tune-content=default ... If you are encoding without the tune, it it not necessary to specify it.
default: tuned for most content. screen: tuned for screen recordings, low complexity animations, and most videos games. Not recommended for complex animation like anime as it disables some post-processing making it look worse in some way.
Forward keyframes and maximum adaptive keyframe distance
... --enable-fwd-kf=1 --kf-max-dist=240 ...
It is recommended to set it to a MAX of 10s worth of frames, or 240 for 24FPS, 300 for 30FPS, and 600 for 60FPS for easier seeking. For video game content or >=60FPS content, is is also possible to use a max of 5s worth of frames to help with seeking performance. Leave it to default if you want maximum efficiency at the cost of worse seeking performance in some footage.
Alternative Reference Noise Reduced frames max number and ARNR strength
... --arnr-maxframes=7 --arnr-strength=5 ... is the default
For --arnr-maxframes: It is recommended to leave it at default unless you want to crank up the efficiency for low motion scenes. For --arnr-strength: It is recommended to leave it at default, although you can lower it to 4 for slightly better noise retention.
Other flags that can be used to give higher efficiency
... --enable-qm=1 --enable-chroma-deltaq=1 --quant-b-adapt=1 --mv-cost-upd-freq=2 ... Description of what each setting does can be found in the description above.
Flags used for most native 10-bit HDR content
... --color-primaries=bt2020 --transfer-characteristics=smpte2084 --matrix-coefficients=bt2020ncl ...
Example command line for great efficiency with good speed for chunked encoding with an 8C16T CPU(1080p24-1080p30 content)
--end-usage=q --cq-level=22 --cpu-used=6 --threads=8 --tile-columns=1 --tile-rows=0 --bit-depth=10 --lag-in-frames=35 --enable-fwd-kf=1 --kf-max-dist=240 --enable-qm=1 --enable-chroma-deltaq=1 --quant-b-adapt=1 --mv-cost-upd-freq=2
Example command line for good efficiency with faster settings and better threading to reduce number of workers needed and RAM consumption
--end-usage=q --cq-level=22 --cpu-used=6 --threads=16 --tile-columns=2 --tile-rows=1 --bit-depth=10 --enable-fwd-kf=1 --kf-max-dist=240