Research map / Video / Video generation
MeanTalker: Efficient and Expressive Speech-Driven 3D Facial Animation via Geometric-Aware Mean Flow
Zhongyuan Zhao, Zhihao Li, Qing Li, Leidong Fan, KANGLIN LIU
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 Mean Flow one-step MeanTalker instead of diffusion MSMD | MeanTalker · TFHP test | LVE: 8.390 -> 7.879 | Improved |
| use MeanTalker instead of DiffPoseTalk for speed | MeanTalker · TFHP test | RTF: 0.919 -> 0.004 | Improved |
| add Geometric-Aware Trajectory Learning (GATL) over MSE-only loss (ablation) | MeanTalker · TFHP test | LVE: 8.735 -> 7.879 | Improved |
| add GATL over MSE-only loss (ablation) | MeanTalker · TFHP test | MOD: 3.187 -> 2.830 | Improved |
| add GATL over MSE-only loss (ablation) | MeanTalker · TFHP test | FDD: 5.152 vs 5.189 | No clear effect |
| add dual-time encoding and rectification for one-step inference (ablation) | MeanTalker · TFHP test | LVE: 9.189 -> 7.879 | Improved |
| add progressive initialization from pretrained standard FM (ablation) | MeanTalker · TFHP test | LVE: 8.272 -> 7.879 | Improved |
| add progressive initialization (convergence speed) (ablation) | MeanTalker · TFHP test | training iterations to converge: approximately 65K vs. 30K iterations | Improved |
| add style code (ablation) | MeanTalker · TFHP test | FDD: 5.919 -> 5.189 | Improved |
| add style code (ablation) | MeanTalker · TFHP test | LVE: 9.144 -> 7.879 | Improved |
| one-step MeanTalker vs 25-step pretrained standard FM (ablation) | MeanTalker · TFHP test | LVE: 8.906 -> 7.879 | Improved |
| MeanTalker vs MSMD in user study lip sync | MeanTalker · TFHP user study | MOS lip sync: 3.69 -> 4.32 | Improved |
Closest papers by meaning
- Probabilistic Speech-Driven 3D Facial Motion Synthesis: New Benchmarks Methods and Applications (CVPR 2024 · similarity 0.91)
- Seeing Speech: Learning Visible Articulatory Dynamics for Speech-Driven 3D Facial Animation (NeurIPS 2026 · similarity 0.90)
- DEEPTalk: Dynamic Emotion Embedding for Probabilistic Speech-Driven 3D Face Animation (AAAI 2025 · similarity 0.89)
- ProsodyTalker: 3D Visual Speech Animation via Prosody Decomposition (AAAI 2025 · similarity 0.89)
- Mimic: Speaking Style Disentanglement for Speech-Driven 3D Facial Animation (AAAI 2024 · similarity 0.88)
- KeyFace: Expressive Audio-Driven Facial Animation for Long Sequences via KeyFrame Interpolation (CVPR 2025 · similarity 0.88)
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.
