Decoupled parallel training of few-step distillation and rollout-based long-horizon adaptation, together with chunk-wise history feature caching, enables real-time 768x512 infinite audio-driven avatars at 27.2 FPS.
Latentsync: Audio conditioned latent diffusion models for lip sync.arXiv e-prints, pages arXiv–2412, 2024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Avatar-Forever: Decoupled Parallel Training for High-Quality Real-Time Infinite Avatars
Decoupled parallel training of few-step distillation and rollout-based long-horizon adaptation, together with chunk-wise history feature caching, enables real-time 768x512 infinite audio-driven avatars at 27.2 FPS.