What it is
Browser demo for real-time talking-head generation from audio, video dubbing, or chat. Streams avatar animation at 30+ FPS on an RTX 3090 using under 12 GB VRAM. Generate and stream talking-head video in the browser from uploaded audio, live mic, source video, or chat responses, with cached avatar registration and preset speaking styles.
Key results
- 30+ FPS on RTX 3090 with under 12 GB GPU memory
- Video dubbing keeps source pose and timing; no motion-reference file needed
- Chat mode streams reply text, TTS audio, and avatar frames incrementally
Shows that browser-based, real-time talking-head generation can unify audio, dubbing, and chat-driven animation in one streaming pipeline.