>>19384
granted, but it still seems bizarre that the pretrained model would use gif input
I would have thought all the compression artifacts would make things harder for the ai and introduce more errors in the model overall
even jpg would be significantly higher quality, at a fraction of the filesize