AI Paper Atlas

Research map / Video / Video generation

Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length

Yubo Huang, Hailong Guo, Fangtai Wu, Weiqiang Wang, Shijie Huang, Qijun Gan, Shifeng Zhang, Lin Liu, Sirui Zhao, Enhong Chen, Jiaming Liu, Steven Hoi

ECCV 2026 · Video › Video generation · approach: Autoregressive long video generation · 58 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
store noisy instead of clean history in the KV cache (History Corrupt) (ablation)Live Avatar · GenBench-LongVideoASE / Dino-S: ASE 2.90 -> 3.42; Dino-S 0.81 -> 0.97Improved
add Adaptive Attention Sink (ablation)Live Avatar · GenBench-LongVideoASE: 3.13 -> 3.42Improved
add Rolling RoPE (ablation)Live Avatar · GenBench-LongVideoDino-S: 0.86 -> 0.97Improved
share the noise level between KV cache and current block (timestep-forcing) vs fixed-noise cache (ablation)Live Avatar · GenBench-LongVideoT.Flicker: 0.891 -> 0.971 (clean cache 0.876)Improved
add Timestep-forcing Pipeline Parallelism on top of DMD (4 GPUs) (ablation)Live Avatar · inference efficiencyFPS: 4.50 (SP4) -> 10.16Improved
apply DMD distillation (80 -> 5 NFE) (ablation)Live Avatar · inference efficiencyFPS: 0.29 -> 3.66Improved
add sequence parallelism across 4 GPUs (ablation)Live Avatar · inference efficiencyFPS: 3.66 -> 4.50No clear effect
add kernel-level optimizations (FP8, FA3, compile) (ablation)Live Avatar · inference efficiencyFPS / TTFF: FPS 20.88 -> 45.2; TTFF 2.89 -> 1.21Improved
use 4 denoising steps instead of 2 (ablation)Live Avatar · GenBench-ShortVideoSync-C: 6.41 -> 7.03Improved
reduce block size from 3 latents to 1 (ablation)Live Avatar · GenBench-ShortVideoSync-C: 7.03 -> 4.12Hurt
increase block size from 3 to 4 latents (ablation)Live Avatar · GenBench-ShortVideoSync-C / TTFF: Sync-C 7.03 -> 6.98; TTFF 0.94 -> 1.21No clear effect
quantize to FP8 (ablation)Live Avatar · GenBench-ShortFPS / IQA: FPS 36.0 -> 45.2; IQA 4.55 -> 4.51Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy