Research map / Video / Video generation
Lip Movements Generation at a Glance
Lele Chen, Zhiheng Li, Ross K Maddox, Zhiyao Duan, Chenliang Xu
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 perceptual loss (ablation) | lip generator · GRID | LMD: 1.33 -> 1.31; SSIM 0.66 -> 0.70; CPBD 0.182 -> 0.209 | Improved |
| add perceptual loss (ablation) | lip generator · GRID | PSNR: 29.66 -> 29.40 | Hurt |
| add correlation loss (ablation) | lip generator, three-stream D. · GRID | LMD: 1.31 -> 1.18; SSIM 0.70 -> 0.73; PSNR 29.40 -> 29.98 | Improved |
| add correlation loss (ablation) | lip generator, two-stream D. · GRID | LMD: 1.42 -> 1.39; SSIM 0.59 -> 0.68 | Improved |
| use derivative of audio and optical flow instead of raw features for correlation (ablation) | correlation loss · GRID | LMD: 1.96 -> 1.18; SSIM 0.52 -> 0.73; PSNR 28.6 -> 29.98 | Improved |
| non-derivative correlation instead of no correlation loss (ablation) | lip generator · GRID | LMD: 1.31 -> 1.96 | Hurt |
| use three-stream discriminator instead of two-stream (ablation) | lip generator with correlation loss · GRID | LMD: 1.39 -> 1.18; SSIM 0.68 -> 0.73; PSNR 29.59 -> 29.98 | Improved |
| use optical flow instead of frame difference as third stream (ablation) | three-stream discriminator · GRID | LMD: 1.40 -> 1.18 | Improved |
| use optical flow instead of frame difference as third stream (ablation) | three-stream discriminator · GRID | CPBD: 0.210 -> 0.175 | Hurt |
| add GAN loss to pixel loss (ablation) | lip generator · GRID | CPBD: 0.006 -> 0.182 | Improved |
| add pixel loss to GAN loss (ablation) | lip generator · GRID | LMD: 1.37 -> 1.33; SSIM 0.67 -> 0.66; CPBD 0.176 -> 0.182 | Improved |
| full model vs Chung et al. (reconstruction loss only) | lip generator · GRID test | LMD: 1.35 -> 1.18 | Improved |
Closest papers by meaning
- SyncTalkFace: Talking Face Generation with Precise Lip-Syncing via Audio-Lip Memory (AAAI 2022 · similarity 0.92)
- Perceptually Accurate 3D Talking Head Generation: New Definitions, Speech-Mesh Representation, and Evaluation Metrics (CVPR 2025 · similarity 0.90)
- OmniSync: Towards Universal Lip Synchronization via Diffusion Transformers (NeurIPS 2025 · similarity 0.89)
- PointTalk: Audio-Driven Dynamic Lip Point Cloud for 3D Gaussian-based Talking Head Synthesis (AAAI 2025 · similarity 0.88)
- Let There Be Sound: Reconstructing High Quality Speech from Silent Videos (AAAI 2024 · similarity 0.88)
- ProsodyTalker: 3D Visual Speech Animation via Prosody Decomposition (AAAI 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.
