SadTalker

SadTalker generates talking head videos from a single portrait and an audio recording. The project states an Apache 2.0 license and removal of its earlier noncommercial restriction. It runs locally on Windows, Linux and macOS, and suits creators who want to animate a face without recording a person on camera. Its animation includes facial expressions and head movement, with examples covering speech and singing in different languages.

The system models facial expression and head pose separately in 3D, then renders them into the video. This approach addresses problems such as unnatural movement, distorted expressions and changes to the person's appearance. Different motion styles can produce different performances from the same recording, and eye blinking is controllable.

SadTalker also supports full-image animation, including full-body portraits, and a still mode for limiting movement. Optional GFPGAN enhancement improves face quality. It saves generated videos as MP4 files.

A local Gradio interface provides browser-based access on your own machine, and a stable-diffusion-webui extension puts portrait animation alongside that image-generation workflow. Local generation uses downloaded model files; an offline GFPGAN patch prevents additional model downloads during generation. Hosted demos on Hugging Face Spaces and Colab run outside your machine. A Discord service also accepts uploaded files and can generate videos from text prompts.