Skip to content

Latest commit

 

History

History
67 lines (57 loc) · 2 KB

README.md

File metadata and controls

67 lines (57 loc) · 2 KB
Raw

metafake

State-of-the-art framework for creating single-shot deepfakes by face swapping, face reenactment, face attribute editing, and face super resolution. Metafake is the first software framework that allows you to combine face swapping and face attribute editing simultaneously. Think of this as RefaceApp+FaceApp on steroids.

Features

Numerous face swapping models that have been optimized and improved

Supported attributes:

  • Hair color
  • Hair style
  • Age
  • More attributes coming soon!

Misc:

  • Complete code re-wrtite. Shrunk code base by orders of magnitude, and made it readable. Elimination of dead code, and optimization of code in general.
  • Created a new, user-friendly API for multiface swapping.
  • Added ability to process an entire folder of destination images/videos.
  • Added ability to swap with multiple-selected models in same program execution
  • Removed temporaily writing images to disk. All image processing is performed in memory.
  • Added config option to output intermediate images into a temporay folder
  • Support for face upscaling models: [GFPGANCleanv1-NoCE-C2|GFPGANv1|GFPGANv1.3|GFPGANv1.4|RestoreFormer|GPEN-BFR-512]
  • Optional replacement of mouth in face mask.

Dependencies

  • python 3.6+
  • pytorch 1.9.0
  • torchvision 0.10.0
  • opencv-python==4.5.2.52
  • pillow==8.0.1
  • numpy
  • imageio== 2.9.0
  • moviepy
  • insightface==0.2.1
  • munch
  • Ninja
  • pytorch_lightning
  • onnxruntime-gpu==1.8.0
  • tqdm==4.51.0
  • keras
  • tensorflow-gpu
  • mxnet
  • scikit-image
  • requests==2.18.4
  • kornia==0.5.4
  • dill
  • wandb
  • onnx==1.9.0

Usage

coming soon

TODO

coming soon

License

coming soon