Genur.art

Using the AbleJones workflow with refinement using Wan Humo. The original = after the head swap, and the other is after passing the head swap video through Wan Humo. If you notice, all the videos generated by the head swap have a kind of ghosting effect, which I believe is due to the displacement of the guide video, because in the inference process it removes the green part with the face and expands the video to 100% of the space, and this leaves these artifacts.