DFAGWE/infinitetalk

InfiniteTalk - Talking Video Generator

Generate realistic talking head videos with accurate lip-sync from images or dub existing videos with new audio!

Features

• —Image-to-Video: Transform a static portrait image into a talking video using audio input
• —Video Dubbing: Re-sync an existing video with new audio while maintaining natural head movements and expressions
• —High Quality: 480p and 720p resolution support with advanced lip-sync technology
• —Unlimited Length: Support for videos of any duration through chunked processing

How It Works

InfiniteTalk uses the state-of-the-art Wan2.1 diffusion model combined with specialized audio conditioning to create photorealistic talking videos. The system synchronizes:

• —Lip movements with audio
• —Head pose and rotations
• —Facial expressions
• —Body posture

Usage

Image-to-Video

1. 1.Upload a portrait image (clear face visibility recommended)
2. 2.Upload an audio file or use the example
3. 3.Adjust parameters if needed
4. 4.Click Generate

Video Dubbing

1. 1.Upload a video with a visible face
2. 2.Upload new audio to dub over it
3. 3.Adjust parameters if needed
4. 4.Click Generate

Parameters

• —Resolution: Choose between 480p (faster) or 720p (higher quality)
• —Diffusion Steps: More steps = higher quality but slower (20-50 recommended)
• —Audio Guide Scale: Controls audio influence on generation (2-4 recommended)
• —Seed: For reproducible results

Credits

Built on InfiniteTalk by MeiGen-AI.

License

Apache 2.0 - See LICENSE.txt for details