/auto-boost-algorithm

Primary LanguagePythonMIT LicenseMIT

Auto-boost Algorithm

Version 1.0: brightness-based

Gets the average brightness of a scene and lowers CQ/CRF/Q the darker the scene is, in a zones.txt file to feed Av1an. The requirements are Vapoursynth, vstools and LSMASHSource.

Usage:

python auto-boost_1.0.py "{animu.mkv}" "{scenes.json}" {base CQ/CRF/Q} "{encoder: aom/svt-av1/rav1e (optional)}"

Example:

python auto-boost_1.0.py "path/to/nice_boat.mkv" "path/to/scenes.json" 30

Advantages:

  • Fast
  • No bs
  • Solves one long-lasting issue of AV1 encoders: low bitrate allocation in dark scenes

Known limitations:

  • Not every dark scene is made equal, brightness is not a great enough metric to determine whether CRF should be decreased or not
  • CRF is boosted to the max during credits
  • Script now entirely irrelevant with SVT-AV1-PSY's new frame-luma-bias feature

Inspiration was drawn from the original Av1an (python) boosting code

Version 2.0: SSIMULACRA2-based

Does a fast encode of the provided file, calculates SSIMULACRA2 scores of each chunks and adjusts CRF per-scene to be closer to the average total score, in a zones.txt file to feed Av1an. The requirements are Vapoursynth, LSMASHSource, fmtconv, Av1an and vapoursynth-ssimulacra2.

Usage:

python auto-boost_2.0.py "{animu.mkv}" {base CQ/CRF/Q}

Example:

python auto-boost_2.0.py "path/to/nice_boat.mkv" 30

Advantages:

  • Lower quality deviation of individual scenes in regards to the entire stream
  • Better allocates bitrate in more complex scenes and compensates by giving less bitrate to scenes presenting some headroom for further compression

Known limitations:

  • Slow process
  • No bitrate cap in place so the size of complex scenes can go out of hand
  • The SSIMULACRA2 metric is not ideal, plus the score alone is not representative enough of if a CRF adjustement is relevant in the context of that scene (AI will save)

Borrowed some code from Sav1or's SSIMULACRA2 script

Version 2.5: SSIMULACRA2&XPSNR-based

The requirements are Vapoursynth, LSMASHSource, Av1an, vszip and a few python library libraries. Optionally: ffmpeg built with XPSNR support, turbo-metrics.

Refined 2.0 with the following additions & changes:

  • Proper argument parsing, more control over the script
  • Separated the fast encode, metric calculation and zone creation into three independant, callable stages
  • Replaced the deprecated vapoursynth-ssimulacra2 by vszip
  • Added turbo-metrics for GPU-accelerated metrics measurement (Nvidia only)
  • Added XPSNR metric and a few zones calculation methods
  • Taking advantage of SVT-AV1-PSY quarter-step CRF feature for more granular control
  • Possibility to use a more aggressive boosting curve
  • And a few other smaller changes...

Many thanks to R1chterScale, Yiss and Kosaka for iterating on auto-boost and making these amazing contributions!

Version 3.0: SSIMULACRA2-based + per-scene grain synthesis strength determination

...and a few other improvements.

Soon:tm: