Paper summary
Paper summary: Loopy: Taming Audio-Driven Portrait Avatar with Long-Term Motion Dependency
Loopy is a video diffusion model that animates a portrait from audio alone, using long-term motion modules instead of the spatial motion templates other methods add to keep movement stable.
- Other audio-driven avatar methods add spatial control signals for stability, which can make motion stiff.
- Loopy uses temporal modules that look across clips to learn natural long-term motion from the data.