Research map / Image generation / Autoregressive and latent generation
UniLS: End-to-End Audio-Driven Avatars for Unified Listening and Speaking
Xuangeng Chu, Ruicong Liu, Yifei Huang, Yun Liu, Yichen Peng, Bo Zheng
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add audio-free stage 1 pretraining of the motion prior (ablation) | UniLS · Seamless Interaction listening | FDD: 25.64 -> 17.12 | Improved |
| add audio-free stage 1 pretraining (ablation) | UniLS · Seamless Interaction listening | F-FID: 5.97 -> 4.304 | Improved |
| add audio-free stage 1 pretraining (ablation) | UniLS · Seamless Interaction speaking | LVE: 6.32 -> 5.83 | Improved |
| add diverse multi-scenario video data to stage 1 (ablation) | UniLS · Seamless Interaction listening | F-FID: 4.62 -> 4.304 | Improved |
| add multi-scenario data to stage 1 (ablation) | UniLS · Seamless Interaction speaking | LVE: 6.26 -> 5.83 | Improved |
| use two separate cross-attention layers for speaker A and B audio instead of one on mixed audio (ablation) | UniLS · Seamless Interaction speaking | LVE: 11.48 -> 5.83 | Improved |
| use dual cross-attention instead of single cross-attention (ablation) | UniLS · Seamless Interaction listening | FDD: 27.46 -> 17.12 | Improved |
| extend ARTalk to speak-listen by adding audio input (ARTalk*) | ARTalk · Seamless Interaction listening | F-FID: 13.143 (DualTalk) vs 10.779 (ARTalk*) | Improved |
| use UniLS end-to-end two-stage training instead of DualTalk | conversational avatar · Seamless Interaction listening | P-FID: 0.079 -> 0.038 | Improved |
| jointly train speaking and listening directly from dual audio without a motion prior (ARTalk*) | ARTalk · Seamless Interaction speaking | PDD: 7.66 -> 8.55 | Hurt |
| use UniLS instead of ARTalk* | conversational avatar · RTX 5090 inference | FPS: 357.7 -> 560.6 | Improved |
Closest papers by meaning
- UniFlow-Audio: Unified Flow Matching for Audio Generation from Omni-Modalities (NeurIPS 2026 · similarity 0.88)
- VASA-1: Lifelike Audio-Driven Talking Faces Generated in Real Time (NeurIPS 2024 · similarity 0.88)
- DAWN: Dynamic Frame Avatar with Non-autoregressive Diffusion Framework for Talking head Video Generation (ICLR 2025 · similarity 0.88)
- OmniTalker: One-shot Real-time Text-Driven Talking Audio-Video Generation With Multimodal Style Mimicking (NeurIPS 2025 · similarity 0.88)
- FaceTalk: Audio-Driven Motion Diffusion for Neural Parametric Head Models (CVPR 2024 · similarity 0.87)
- GAIA: Zero-shot Talking Avatar Generation (ICLR 2024 · similarity 0.87)
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.
