with torch.no_grad(): fake_frames = model(face_sequences, audio_features)
Most implementations require a configuration file (like vox-adv-256.yaml ) that tells the code how to read the checkpoint.
Understanding vox-adv-cpk.pth.tar: The Engine Behind Real-Time Face Animation
Processing multiple frames simultaneously can overload your VRAM. If you experience crashes, lower your batch size to 1 or run the inference on CPU (though it will be significantly slower). Ethical Considerations
The model uses the weights inside Vox-adv-cpk.pth.tar to automatically detect facial landmarks (eyes, mouth, jawline) on both the source image and the driving video without any prior manual labeling.