A set of AI models is trained targeting different deepfake techniques with different custom made dataset with pytorch. We explored alternatives such as TensorFlow and Keras but chose PyTorch for its flexibility and dynamic computation graph.

Best for deeplearning and AI

When you're training a SOTA generative model with two people the giant wealth of existing GitHub Issues threads/reference codebases/etc is essential.

ML hardware acceleration using CUDA for Nvidia GPUs on Windows, and via MPS and MLX frameworks for Apple silicon on macOS!

Empowers project with efficient model development and deployment, including ONNX integration.