Hallo

What Hallo is

Hallo generates an animated portrait video from a still image and driving speech audio. Its audio-driven synthesis is designed to coordinate lip movement, facial expression and pose with the sound, and inference options provide separate weights for pose, face and lips. The input portrait should be square, with a forward-facing face filling 50%–70% of the image and rotated less than 30 degrees. Driving audio must be a clear-vocal English WAV file; background music is acceptable. The default output is an MP4 saved to `.cache/output.mp4`, though users can choose another path. Hallo is free and open source under an MIT license. Its listed system requirements are Ubuntu 20.04 or 22.04 and CUDA 12.1, and the project says it tested the software on A100 GPUs. Pretrained checkpoints and supporting models are available through its Hugging Face repository. The project also provides training and data-preprocessing code, including multi-node and multi-GPU support, and links community resources for Windows, ComfyUI, a WebUI, Docker and RunPod. Its project materials identify deepfake misuse, privacy and consent as risks and call for responsible use.