New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
HandBrake is changing the default SVT-AV1 tune #6012
Comments
"None" is not "Default". HandBrake's psnr tune is 1, i.e. "default". There aren't really "defaults" in HandBrake. There are presets instead which define settings for particular use cases. However, "None" may be a poor name here since 0 is "VQ" in svt-av1 parlance. All the names are pretty cryptic for someone not familiar with encoder jargon. |
That is of course correct. My concern was rather that if you select the SVT-AV1 encoder without adjusting any other settings, then I think HandBrake should also use all the default settings SVT-AV1 is using. |
Generally speaking, HandBrake doesn't reset to default values when changing the encoder. HandBrake will not change any of the other settings unless the current option for some setting doesn't exist for the new encoder. It would be quite annoying if you set up several settings, then noticed you had neglected to set the encoder first, changed the encoder, and had to go back and set all your settings again. |
Hmm, I seem to be mistaken in the above assertion. It is changing more than I thought when changing encoders. Given the current behavior, it does seem like it should be picking a reasonable default when changing values. I'm not sure if tune "None" is the actual default for any of the other encoders supported by HandBrake. I'm pretty sure it's not been a priority to try to align with encoder application defaults. They are often a moving target. |
I might add that encoder application defaults are sometimes not good defaults. In particular, tuning for psnr is generally not used as a default in other encoders because an encode with the best psnr is not necessarily an encode that is visually the best looking. Encoders generally employ psycho-visual optimizations that do not produce the best psnr (or ssim for that matter). |
I was told that SVT-AV1 did choose to set PSNR as the default tune because "it's the happy all objective metrics mode and that has been the focus of the team". |
Like jstebbins said, defaults are set by presets, and we mostly don't reset anything when switching encoder. The GTK UI might be resetting the quality value too, the Mac GUI doesn't. I originally named it "none" to keep it similar to other encoders. We could rename it if it doesn't break anything in the UI, but I wouldn't go all the way to make it automatically change to "psnr" when svt-av1 is manually selected, unless VQ is terribly broken. |
psnr and ssim tunes are typically used by developers to get a metric for a particular change they are making to measure it's effect. So it's not really too surprising that an encoder that has been under heavy development would have this as a default. I imagine they would be using ssim if it were not experimental since it is usually a better metric when it comes to perceived visual quality. Maybe at this point in development psnr does have some advantages given it's what the developers use. But typically in a mature encoder it's strictly used for getting metrics in a testing scenario. |
Seems it not only confuses me: https://www.reddit.com/r/AV1/comments/14m04vg/is_handbrakes_svtav1_encoder_tune_none_same_as/ |
Problem Description
Not sure if this is by purpose. But HandBrake is changing the default tune setting for SVT-AV1.
According to the documentation, the default setting for
--tune
is 1:https://gitlab.com/AOMediaCodec/SVT-AV1/-/blob/master/Docs/Parameters.md
But if "None" is selected for the tune for the SVT-AV1 encoder, 0 is used.
Which is set here:
HandBrake/libhb/encsvtav1.c
Line 197 in b9e531f
Activity Log, Crash Log or any other details
What Operating System are you running?
All
What version of HandBrake are you running?
Current Nightly
Where did you download HandBrake from?
No response
The text was updated successfully, but these errors were encountered: