preprint · arXiv preprint, 2026 · Aug 12, 2026
Avatar-Forever: Decoupled Parallel Training for High-Quality Real-Time Infinite Avatars
Ruibin Li, Zhiyuan Ma, Fangzhou Ai, Lei Zhang
In plain English
Proposes a framework for streaming audio-driven avatar video of unlimited length by training a fast generator and a stability adapter in parallel, plus a caching method to cut computation. The authors report about 27 frames per second at 768x512 on a single H100 GPU.
real-time avatars
talking head generation
latency