AvatarForcing: One-Step Streaming Talking Avatars via Local-Future Sliding-Window Denoising
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910057087631360 |
|---|---|
| author | Cui, Liyuan Hu, Wentao Zhang, Wenyuan Yang, Zesong Shi, Fan Liu, Xiaoqiang |
| author_facet | Cui, Liyuan Hu, Wentao Zhang, Wenyuan Yang, Zesong Shi, Fan Liu, Xiaoqiang |
| contents | Real-time talking avatar generation requires low latency and minute-level temporal stability. Autoregressive (AR) forcing enables streaming inference but suffers from exposure bias, which causes errors to accumulate and become irreversible over long rollouts. In contrast, full-sequence diffusion transformers mitigate drift but remain computationally prohibitive for real-time long-form synthesis. We present AvatarForcing, a one-step streaming diffusion framework that denoises a fixed local-future window with heterogeneous noise levels and emits one clean block per step under constant per-step cost. To stabilize unbounded streams, the method introduces dual-anchor temporal forcing: a style anchor that re-indexes RoPE to maintain a fixed relative position with respect to the active window and applies anchor-audio zero-padding, and a temporal anchor that reuses recently emitted clean blocks to ensure smooth transitions. Real-time one-step inference is enabled by two-stage streaming distillation with offline ODE backfill and distribution matching. Experiments on standard benchmarks and a new 400-video long-form benchmark show strong visual quality and lip synchronization at 34 ms/frame using a 1.3B-parameter student model for realtime streaming. Our page is available at: https://cuiliyuan121.github.io/AvatarForcing/ |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_14331 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AvatarForcing: One-Step Streaming Talking Avatars via Local-Future Sliding-Window Denoising Cui, Liyuan Hu, Wentao Zhang, Wenyuan Yang, Zesong Shi, Fan Liu, Xiaoqiang Computer Vision and Pattern Recognition Real-time talking avatar generation requires low latency and minute-level temporal stability. Autoregressive (AR) forcing enables streaming inference but suffers from exposure bias, which causes errors to accumulate and become irreversible over long rollouts. In contrast, full-sequence diffusion transformers mitigate drift but remain computationally prohibitive for real-time long-form synthesis. We present AvatarForcing, a one-step streaming diffusion framework that denoises a fixed local-future window with heterogeneous noise levels and emits one clean block per step under constant per-step cost. To stabilize unbounded streams, the method introduces dual-anchor temporal forcing: a style anchor that re-indexes RoPE to maintain a fixed relative position with respect to the active window and applies anchor-audio zero-padding, and a temporal anchor that reuses recently emitted clean blocks to ensure smooth transitions. Real-time one-step inference is enabled by two-stage streaming distillation with offline ODE backfill and distribution matching. Experiments on standard benchmarks and a new 400-video long-form benchmark show strong visual quality and lip synchronization at 34 ms/frame using a 1.3B-parameter student model for realtime streaming. Our page is available at: https://cuiliyuan121.github.io/AvatarForcing/ |
| title | AvatarForcing: One-Step Streaming Talking Avatars via Local-Future Sliding-Window Denoising |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.14331 |