SvtAv1EncApp already has PSNR & SSIM RDO tunes, so I think an XPSNR tune would go a very long way. I am definitely in over my head with this one though, so let me know if you can help, or just make a PR!

It would be nice to have GUI or a how-to for some slightly less involved in the scene people to avoid spending lots of time figuring things out. Otherwise, thanks for your involvement! :-)

• lavish branch is a "testing" branch (should probably be renamed as such) & other branches should be testing1, testing2, etc
• document recent changes to master branch each time they are committed - those making commits should remember to do this
• changes that are merged to master have to be proved metrically and psychovisually through heuristic evaluation. ideally one must test XPSNR and SSIMU2, but an SSIMU2 + visual improvement will suffice.

Settings: --variance-boost-strength 3 --new-variance-octile 4 --crf 40

Parasite @ 39:22, Blu-Ray source
There's incoming sunlight light coming from the car's back window that's shining through the passenger's face and hair. The hair shows noticeable pulsing (detail disappearing and reappearing), which doesn't happen in the 64x64 version of the encode.

Avenues to explore:

• Find how variance in the affected superblocks varies over time
• Explore making the selected boost from 8x8 variances more robust (averaging between values near octile, temporal smoothing, etc)

Similar to aom-av1-lavish's luma bias, it'd be a great feature to have in SVT-AV1-PSY

SVT-AV1 currently doesn't have a film grain/noise estimator in its code, so having one will help unlock working on these features:

• Improve performance of variance boost with heavy noise and film grain content
• --film-grain auto setting