Changing from normal SVT-AV1 to SVT-AV1-PSY

[BlueSwordM]

This change serves to improve the visual quality of the video output created by the encoder, as well as improving the its defaults.

I will be the main maintainer for this package as I'm always on the lookout for changes and whatnot.

This will allow a large differentiating factor for CachyOS.

[lseman]

I made some comments, but overall LGTM.

[gianni-rosato]

Hi, just wanted to give my two cents as the maintainer of SVT-AV1-PSY. First of all, I'm super excited to see this being discussed here!

For those reviewing the PR, some info about SVT-AV1-PSY:

• As you can easily discover, our encoder is focused on being the best AV1 encoder for perceptual fidelity. The reason for this being a fork is that mainline SVT-AV1 has different priorities not necessarily perfectly aligned with maximum visual quality. That being said, we don't intend to compromise on speed or stability to achieve our goals, and so far we can assure we haven't.
• We cannot ensure the same degree of stability as mainline can, but we can get very close. The master branch of SVT-AV1-PSY features automated build testing for every commit, and our release cycle is aligned with mainline's (excluding release candidates). We are in touch with a sizeable community of encoders who are constantly using SVT-AV1-PSY and are quick to discover any regressions in the testing branch, so things should always be stable upon reaching master and afterwards being released.
• SVT-AV1-PSY is a superset of SVT-AV1, so any string of command line arguments passed to mainline SvtAv1EncApp should work with our encoder (the binary is of the same name, too). That being said, a command passed to our encoder may not work with mainline due to additional PSY options (e.g --tune 3). We won't make any guarantees that bitstreams produced by our encoder and mainline using the same parameters will be identical.

For any others reviewing this PR, I hope this clears things up as how SVT-AV1-PSY would integrate into CachyOS!

[ptr1337]

Could we please provide this is an own version instead of replacing the default stable svt-av1?
I dont think it is good to replace the normal upstream release, with a custom one.

Besides that it is looking fine to me :)

[ptr1337]

Hi, just wanted to give my two cents as the maintainer of SVT-AV1-PSY. First of all, I'm super excited to see this being discussed here!

For those reviewing the PR, some info about SVT-AV1-PSY:

• As you can easily discover, our encoder is focused on being the best AV1 encoder for perceptual fidelity. The reason for this being a fork is that mainline SVT-AV1 has different priorities not necessarily perfectly aligned with maximum visual quality. That being said, we don't intend to compromise on speed or stability to achieve our goals, and so far we can assure we haven't.
• We cannot ensure the same degree of stability as mainline can, but we can get very close. The master branch of SVT-AV1-PSY features automated build testing for every commit, and our release cycle is aligned with mainline's (excluding release candidates). We are in touch with a sizeable community of encoders who are constantly using SVT-AV1-PSY and are quick to discover any regressions in the testing branch, so things should always be stable upon reaching master and afterwards being released.
• SVT-AV1-PSY is a superset of SVT-AV1, so any string of command line arguments passed to mainline SvtAv1EncApp should work with our encoder (the binary is of the same name, too). That being said, a command passed to our encoder may not work with mainline due to additional PSY options (e.g --tune 3). We won't make any guarantees that bitstreams produced by our encoder and mainline using the same parameters will be identical.

For any others reviewing this PR, I hope this clears things up as how SVT-AV1-PSY would integrate into CachyOS!

Thanks gianni!
That did clear things up, but as i have written above, i dont think it is good to replace the default stable version with a -git version.

I really want to avoid using git release in cachyos as default and also replacing offical (maintained) versions of software, with a less tested custom software.
Providing an additonal software, which user can install/replacing them on their own is a better solution.

[BlueSwordM]

@vnepogodin I pushed the package name change as requested.

If @ptr1337 accepts, I believe we can push it?
I'm still not sure if it's not better to close this PR and make a new PR with a new PKGBUILD.

That way, the package only gets updated when svt-av1-psy gets updated, rather than with any and every svt-av1 update.