The folks at Disney Research Studios have moved one step closer to replacing human actors by developing this algorithm for high resolution face-swapping in images and videos.
To the best of our knowledge, this is the first method capable of rendering photo-realistic and temporally coherent results at megapixel resolution. To this end, we introduce a progressively trained multi-way comb network and a light- and contrast-preserving blending method. We also show that while progressive training enables generation of high-resolution images, extending the architecture and training data beyond two people allows us to achieve higher fidelity in generated expressions. When compositing the generated expression onto the target face, we show how to adapt the blending strategy to preserve contrast and low-frequency lighting. Finally, we incorporate a refinement strategy into the face landmark stabilization algorithm to achieve temporal stability, which is crucial for working with high-resolution videos. We conduct an extensive ablation study to show the influence of our design choices on the quality of the swap and compare our work with popular state-of-the-art methods.
This is basically like supercharged DeepFakes and the results are pretty impressive. Disney Research Studios claims they're developing this for movies and visual effects, but I can think of some other *ahem* more adult oriented applications as well. Yes, I'm talking about porn. What else would I be talking about? I'm always talking about porn. This is the Internet. Everybody's always talking about porn. Why is Disney Research Studios even pretending they're not doing this for porn? I wonder how many more times I can type Disney and porn in the same sentence before I get sued.
Keep going for a video of the face-swapping in action. It's still not perfect, but it's significantly better than the alternative methods.