Skip to content

Cross-platform command-line AV1 / VP9 / HEVC / VVC encoding framework with per scene quality encoding

License

Notifications You must be signed in to change notification settings

SebastianGode/Av1an

 
 

Repository files navigation


Av1an

A cross-platform framework to streamline encoding

alt text

Easy, Fast, Efficient and Feature Rich

An easy way to start using VVC / AV1 / HEVC / VP9 / VP8 encoding. AOM, RAV1E, SVT-AV1, SVT-VP9, VPX, x265, VTM are supported.

Example with default parameters:

av1an -i input

With your own parameters:

av1an -i input -enc aom -v " --cpu-used=3 --end-usage=q --cq-level=30 --threads=8 " -w 10
--split_method aom_keyframes --vmaf_target 95 --vmaf_path "vmaf_v0.6.1.pkl" -min_q 20 -max_q 60
-ff " -vf scale=-1:1080 "  -a " -c:a libopus -ac 2 -b:a 192k " -s scenes.csv -log my_log -o output

Usage

-i   --file_path        Input file(s) (relative or absolute path). Will be processed with same
                        settings.

-o   --output_file      Name/Path for output file (Default: (input file name)_av1.mkv)
                        Output file ending is always `.mkv`

-enc --encoder          Encoder to use (`aom`,`rav1e`,`svt_av1`,`svt_vp9`,`vpx`,`x265`,`vvc`)
                        Default: aom
                        Example: -enc rav1e

-v   --video_params     Encoder settings flags (If not set, will be used default parameters.)
                        Must be inside ' ' or " "

-p   --passes           Set number of passes for encoding
                        (Default: AOMENC: 2, rav1e: 1, SVT-AV1: 1, SVT-VP9: 1, 
                        VPX: 2, x265: 1, x264: 1, VVC:1)

-w   --workers          Override number of workers.
                            
--resume                If encode was stopped/quit resumes encode with saving all progress
                        Resuming automatically skips scenedetection, audio encoding/copy,
                        spliting, so resuming only possible after actuall encoding is started.
                        /.temp folder must be presented for resume.

--no_check              Skip checking numbers of frames for source and encoded chunks.
                        Needed if framerate changes to avoid console spam.
                        By default any differences in frames of encoded files will be reported.

--keep                  Not deleting temprally folders after encode finished.

-log --logging          Path to .log file(By default created in temp folder)

--temp                  Set path for temporally folders. Default: .temp

-cfg                    Save/Read config file with encoder, encoder parameters,
                        FFmpeg and audio settings.

--mkvmerge              Use mkvmerge for concatenating instead of ffmpeg.
                        Files will be a bit smaller. Recommended for x265.

FFmpeg options

-a   --audio_params     FFmpeg audio settings (Default: copy audio from source to output)
                        Example: -a '-c:a libopus -b:a  64k'

-ff  --ffmpeg           FFmpeg options. Applied to each segment individually.
                        Example:
                        --ff " -vf scale=320:240 "

-fmt --pix_format       Setting custom pixel/bit format for piping
                        (Default: 'yuv420p')
                        Example for 10 bit source: 'yuv420p10le'
                        Based on encoder, options should be adjusted accordingly.

Segmenting

--split_method          Method used for generating splits.(Default: PySceneDetect)
                        Options: `pyscene`, `aom_keyframes`
                        `pyscene` - PyScenedetect, content based scenedetection
                        with threshold.
                        `aom_keyframes` - using stat file of 1 pass of aomenc encode
                        to get exact place where encoder will place new keyframes.
                        (Keep in mind that speed also depends on set aomenc parameters)

-cm  --chunk_method     Determine way in which chunks made for encoding.
                        ['hybrid'(default), 'select', 'vs_ffms2']

-tr  --threshold        PySceneDetect threshold for scene detection Default: 35

-s   --scenes           Path to file with scenes timestamps.
                        If given `0` spliting will be ignored
                        If file not exist, new will be generated in current folder
                        First run to generate stamps, all next reuse it.
                        Example: "-s scenes.csv"

-xs  --extra_split      Adding extra splits if frame distance beetween splits bigger than
                        given value. Works with/without PySceneDetect
                        Example: 1000 frames video with single scene,
                        -xs 200 will add splits at 200,400,600,800.

Target VMAF

--vmaf_target           Vmaf value to target. Supported for all encoders(Exception:VVC).
                        Best works in range 85-97.
                        When using this mode specify full encoding options.
                        Encoding options must include quantizer based mode,
                        and some quantizer option provided. (This value got replaced)
                        `--crf`,`--cq-level`,`--quantizer` etc
                        
--min_q, --max_q        Min,Max Q values limits for Target VMAF
                        If not set by user, encoder default will be used.
                        
--vmaf                  Calculate vmaf after encode is done.
                        showing vmaf values for all frames,
                        mean, 1,25,75 percentile.
                        
--vmaf_plots            Make plots for target_vmaf search decisions
                        (Exception: early skips)
                        Saved in temp folder

--vmaf_path             Custom path to libvmaf models.
                        example: --vmaf_path "vmaf_v0.6.1.pkl"
                        Recomended to place both files in encoding folder
                        (`vmaf_v0.6.1.pkl` and `vmaf_v0.6.1.pkl.model`)
                        (Required if vmaf calculation doesn't work by default)

--vmaf_res              Resolution scaling for vmaf calculation, 
                        vmaf_v0.6.1.pkl is 1920x1080 (by default),
                        vmaf_4k_v0.6.1.pkl is 3840x2160 (don't forget about vmaf_path)

--vmaf_steps            Number of probes for interpolation.
                        1 and 2 probes have special cases to try to work with few data points.
                        Optimal is 4-6 probes. Default: 4

--vmaf_rate             Setting rate for vmaf probes (Every N frame used in probe, Default: 4)

--n_threads             Limit number of threads that used for vmaf calculation
                        Example: --n_threads 12
                        (Required if VMAF calculation gives error on high core counts)

Main Features

Spliting video by scenes for parallel encoding because AV1 encoders are currently not good at multithreading, encoding is limited to single or couple of threads at the same time.

  • PySceneDetect used for spliting video by scenes and running multiple encoders.
  • Fastest way to encode AV1 without losing quality, as fast as many CPU cores you have :).
  • Target VMAF mode. Targeting end result visual quality.
  • Resuming encoding without loss of encoded progress.
  • Simple and clean console look.
  • Automatic detection of the number of workers the host can handle.
  • Building encoding queue with bigger files first, minimizing waiting for the last scene to encode.
  • Both video and audio transcoding with FFmpeg.
  • Logging of progress of all encoders.
  • "Boosting" quality of dark scenes based on their brightness.

Install

Support developer

Bitcoin - 1gU9aQ2qqoQPuvop2jqC68JKZh5cyCivG

About

Cross-platform command-line AV1 / VP9 / HEVC / VVC encoding framework with per scene quality encoding

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 98.7%
  • Rust 1.3%