Research map / Video / Video generation
VoCa: Unified Autoregressive Modeling for Talking Audio-Video Generation
Zhuofan Zong, Jiale Yuan, Yufei Liu, Dongzhi Jiang, Hao Shao, Zimu Lu, Ke Wang, Yunqiao Yang, Mingjie Zhan, Hongsheng LI
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 |
|---|---|---|---|
| use both LLM audio tokens and hidden states instead of tokens only (ablation) | VoCa-1.3B · HDTF (VoCa-1.3B ablation) | FVD: 59.961 -> 39.630 | Improved |
| use both tokens and hidden states instead of hidden states only (ablation) | VoCa-1.3B · HDTF (VoCa-1.3B ablation) | Sync-C: 3.773 -> 6.082 | Improved |
| add auxiliary HuBERT distillation loss (lambda1 = 1) (ablation) | VoCa-1.3B · HDTF (VoCa-1.3B ablation) | Sync-C: 4.782 -> 6.082 | Improved |
| raise distillation weight from 1 to 2 (ablation) | VoCa-1.3B · HDTF (VoCa-1.3B ablation) | FVD: 39.630 -> 64.509 | Hurt |
| add 6-layer conv speech refiner instead of none (ablation) | VoCa-1.3B · HDTF (VoCa-1.3B ablation) | FVD: 47.177 -> 39.630 | Improved |
| deepen refiner from 6 to 8 layers (ablation) | VoCa-1.3B · HDTF (VoCa-1.3B ablation) | FVD: 39.630 -> 39.269 | No clear effect |
| scale video backbone from Wan2.1-1.3B to Wan2.2-5B (ablation) | VoCa · HDTF | Sync-C: 6.082 -> 8.054 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Hallo3 — Hallo3: Highly Dynamic and Realistic Portrait Image Animation with Video Diffusion Transformer (CVPR 2025) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- OmniTalker: One-shot Real-time Text-Driven Talking Audio-Video Generation With Multimodal Style Mimicking (NeurIPS 2025 · similarity 0.89)
- DAWN: Dynamic Frame Avatar with Non-autoregressive Diffusion Framework for Talking head Video Generation (ICLR 2025 · similarity 0.88)
- AudioCALM: Continuous Autoregressive Language Modeling for Universal Audio Generation (NeurIPS 2026 · similarity 0.88)
- Native Audio-Visual Alignment for Generation (NeurIPS 2026 · similarity 0.88)
- Tell What You Hear From What You See - Video to Audio Generation Through Text (NeurIPS 2024 · similarity 0.87)
- Continuous Autoregressive Modeling with Stochastic Monotonic Alignment for Speech Synthesis (ICLR 2025 · 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.
