Research map / Video / Video generation
FLOAT: Generative Motion Latent Flow Matching for Audio-driven Talking Portrait
Taekyung Ki, Dongchan Min, Gyeongsu Chae
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 frame-wise AdaLN with masked self-attention instead of cross-attention (ablation) | FLOAT · HDTF | E-FID: 1.452 -> 1.229 | Improved |
| use frame-wise AdaLN instead of cross-attention (ablation) | FLOAT · HDTF | LSE-D: 7.757 -> 7.290 | Improved |
| use flow matching instead of epsilon-prediction diffusion (ablation) | FLOAT · HDTF | LSE-D: 9.922 -> 7.290 | Improved |
| use flow matching instead of x0-prediction diffusion (ablation) | FLOAT · HDTF | LSE-D: 9.048 -> 7.290 | Improved |
| use flow matching (10 NFE) instead of diffusion (50 steps) (ablation) | FLOAT · HDTF | FID: 21.190 -> 21.100 | No clear effect |
| add facial component perceptual loss to motion auto-encoder (ablation) | motion auto-encoder · HDTF | FID: 21.061 -> 19.803 | Improved |
| add facial component perceptual loss to motion auto-encoder (ablation) | motion auto-encoder · RAVDESS | FID: 28.866 -> 23.350 | Improved |
| increase NFE from 2 to 10 (ablation) | FLOAT · HDTF | FVD: 178.831 -> 162.052 | Improved |
| increase NFE from 10 to 20 (ablation) | FLOAT · HDTF | FVD: 162.052 -> 164.392 | No clear effect |
| increase audio guidance scale from 1 to 2 (ablation) | FLOAT · RAVDESS | LSE-D: 7.049 -> 6.994 | Improved |
| increase emotion guidance scale from 1 to 2 (audio scale 1) (ablation) | FLOAT · RAVDESS | E-FID: 1.555 -> 1.334 | Improved |
| add 3DMM head pose condition | FLOAT · HDTF | FVD: 162.052 -> 126.663 | 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.
- EdTalk — EDTalk: Efficient Disentanglement for Emotional Talking Head Synthesis (ECCV 2024) · better on 1 of 1 shared result
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
- GoHD: Gaze-oriented and Highly Disentangled Portrait Animation with Rhythmic Poses and Realistic Expressions (AAAI 2025 · similarity 0.90)
- Teller: Real-Time Streaming Audio-Driven Portrait Animation with Autoregressive Motion Generation (CVPR 2025 · similarity 0.89)
- Hallo3: Highly Dynamic and Realistic Portrait Image Animation with Video Diffusion Transformer (CVPR 2025 · similarity 0.89)
- DAWN: Dynamic Frame Avatar with Non-autoregressive Diffusion Framework for Talking head Video Generation (ICLR 2025 · similarity 0.89)
- Hallo2: Long-Duration and High-Resolution Audio-Driven Portrait Image Animation (ICLR 2025 · similarity 0.89)
- Loopy: Taming Audio-Driven Portrait Avatar with Long-Term Motion Dependency (ICLR 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.
