Research map / Video / Video generation
Emotion-Conditioned Motion Sub-spaces with Flow Matching for Real-Time Audio-Driven Talking Heads
Haoyu Wang, Xiaozhe Xin, Xiaoyu Qin, Meiguang Jin, Junfeng Ma, Dan Xu, Jia Jia
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 |
|---|---|---|---|
| Remove emotion-conditioned sub-spaces (use motion dictionary directly) (ablation) | Ours · MEAD | emotion accuracy (%): 82.05 -> 64.10 | Improved |
| Choose random bases instead of learned emotion sub-spaces (ablation) | Ours · MEAD | emotion accuracy (%): 82.05 -> 45.00 | Hurt |
| Use learned emotion sub-spaces vs random bases (ablation) | Ours · MEAD | FVD: 84.30 -> 50.39 | Improved |
| Use layer-progressive cross-attention fusion instead of naive concatenation (ablation) | Ours · MEAD | emotion accuracy (%): 33.33 -> 82.05 | Improved |
| Use layer-progressive fusion instead of self-attention fusion (ablation) | Ours · MEAD | FVD: 64.30 -> 50.39 | Improved |
| Use layer-progressive fusion instead of cross-attention in all layers (ablation) | Ours · MEAD | emotion accuracy (%): 69.44 -> 82.05 | Improved |
| Use layer-progressive fusion instead of cross-attention in all layers (ablation) | Ours · MEAD | LSE-C: 4.34 -> 7.52 | Improved |
| Ten-step flow-matching sampler with motion dictionary instead of 40-step diffusion (Hallo) | Ours · inference speed | FPS: 0.55 -> 91.23 | Improved |
| Ours vs diffusion baselines on emotion accuracy | Ours vs Hallo/EchoMimic/Sonic/FLOAT · RAVDESS | emotion accuracy (%): 45.00 (best baseline FLOAT) -> 62.50 | Improved |
| Ours vs baselines on LSE-C lip-sync confidence | Ours · MEAD | LSE-C: 7.92 vs EAT 7.76, EDTalk 7.95, Sonic 8.09 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Sonic — Sonic: Shifting Focus to Global Audio Perception in Portrait Animation (CVPR 2025) · better on 4 of 4 shared results
- EDTalk — EDTalk: Efficient Disentanglement for Emotional Talking Head Synthesis (ECCV 2024) · better on 3 of 3 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
- DEEPTalk: Dynamic Emotion Embedding for Probabilistic Speech-Driven 3D Face Animation (AAAI 2025 · similarity 0.91)
- FaceTalk: Audio-Driven Motion Diffusion for Neural Parametric Head Models (CVPR 2024 · similarity 0.91)
- Style2Talker: High-Resolution Talking Head Generation with Emotion Style and Art Style (AAAI 2024 · similarity 0.90)
- KeyFace: Expressive Audio-Driven Facial Animation for Long Sequences via KeyFrame Interpolation (CVPR 2025 · similarity 0.90)
- EmotiveTalk: Expressive Talking Head Generation through Audio Information Decoupling and Emotional Video Diffusion (CVPR 2025 · similarity 0.90)
- GoHD: Gaze-oriented and Highly Disentangled Portrait Animation with Rhythmic Poses and Realistic Expressions (AAAI 2025 · similarity 0.90)
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.
