Replies: 1 comment
-
Well, that lowres output was just worked around by the subsampling method invented by hillobar from rope. It works good enough for me. |
Beta Was this translation helpful? Give feedback.
-
Well, that lowres output was just worked around by the subsampling method invented by hillobar from rope. It works good enough for me. |
Beta Was this translation helpful? Give feedback.
-
It seems the entire face swapper community is held back by the inswapper_128 low res model. I was looking at facechain and I noticed they seem to be using a different model for the swapping portion. Can any benefit be derived for use in Roop Unleashed? As far as I can tell (through machine translation) this model is open source.
Here's a link: https://www.modelscope.cn/models/iic/cv_unet_face_fusion_torch/summary
Beta Was this translation helpful? Give feedback.
All reactions