Face Restore Codeformer / GPFGAN slow speed compared to A1111 #311
TheRealPhooney
started this conversation in
General
Replies: 3 comments
-
Verified. We had to disable face restore unittests in CQ as they just get stuck there. |
Beta Was this translation helpful? Give feedback.
0 replies
-
Beta Was this translation helpful? Give feedback.
0 replies
-
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
I couldn't see anybody else talking about this, so I wanted to check if it was just me.
Whenever I use face restore, either as part of txt2img/img2img or within the Reactor extension, the face restore part seems to take a lot longer than it did on A1111.
A quick and dirty comparison is a 512x768 image taking 3-4 seconds without any face restoration, and 12-14 seconds with face restoration, so 9-11 seconds for the GPFGAN/Codeformer to do its thing.
It's especially noticeable if you were using AnimateDiff, resulting in a bunch of frames to process.
Unfortunately, I pretty much gutted my A1111 installation moving all kinds of stuff over to Forge, and it would be a bit of a drag for me to set up A1111 again so all I have is my "feeling" that this part of the process is a lot slower rather than a direct comparison.
Is anybody else finding this slowdown too?
Beta Was this translation helpful? Give feedback.
All reactions