warmshao/FasterLivePortrait
Real-time portrait animation system that drives static images into animated video using optimized deep learning inference.

FasterLivePortrait is an optimized implementation of LivePortrait that achieves real-time portrait animation on GPU hardware. It leverages TensorRT acceleration to reach 30+ FPS performance including pre- and post-processing. The system supports driving portraits from video, audio, or text input, with the ability to animate multiple faces simultaneously and even animate animal portraits. It provides a native Gradio interface and ONNX/TensorRT deployment options.