Research map / Video / Video generation
Implicit Identity Representation Conditioned Memory Compensation Network for Talking Head video Generation
Fa-Ting Hong, Dan 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 memory compensation module (MCM) to baseline (ablation) | baseline talking head model · VoxCeleb1 same-identity | SSIM (%): 81.1 -> 82.3 | Improved |
| add IICM identity-conditioned memory on top of MCM (ablation) | Baseline + MCM · VoxCeleb1 same-identity | SSIM (%): 82.3 -> 82.5 | Improved |
| use channel split in MCM (ablation) | MCM · VoxCeleb1 same-identity | SSIM (%): 82.0 -> 82.3 | Improved |
| plug IICM and MCM into FOMM | FOMM · VoxCeleb1 same-identity | SSIM (%): 72.3 -> 81.8 | Improved |
| plug IICM and MCM into TPSN | TPSN · VoxCeleb1 same-identity | SSIM (%): 81.6 -> 82.0 | Improved |
| apply positional encoding to keypoints with L=20 (ablation) | MCNet · VoxCeleb1 same-identity | SSIM (%): 82.5 -> 69.4 | Hurt |
| apply positional encoding with L=10 to keypoints (ablation) | MCNet · VoxCeleb1 same-identity | SSIM (%): 82.5 -> 82.4 | No clear effect |
| deploy IICM and MCM in all layers rather than only the first (ablation) | MCNet · VoxCeleb1 same-identity | SSIM (%): 82.3 -> 82.5 | Improved |
| change memory bank spatial size from 32 to 16 or 64 (ablation) | MCNet · VoxCeleb1 same-identity | SSIM (%): 82.5 -> 82.3 (16), 82.1 (64) | No clear effect |
| remove consistency regularization on memory | MCNet · meta memory visualization | memory pattern quality | Hurt |
| test on identities seen in training | MCNet · HDTF-based split | SSIM (%): 82.5 -> 83.6 | Improved |
Closest papers by meaning
- High-Fidelity and Freely Controllable Talking Head Video Generation (CVPR 2023 · similarity 0.89)
- Synergizing Motion and Appearance: Multi-Scale Compensatory Codebooks for Talking Head Video Generation (CVPR 2025 · similarity 0.88)
- Memories are One-to-Many Mapping Alleviators in Talking Face Generation (TPAMI 2024 · similarity 0.88)
- Occlusion-Insensitive Talking Head Video Generation via Facelet Compensation (AAAI 2025 · similarity 0.88)
- FaceChain-ImagineID: Freely Crafting High-Fidelity Diverse Talking Faces from Disentangled Audio (CVPR 2024 · similarity 0.87)
- FG-EmoTalk: Talking Head Video Generation with Fine-Grained Controllable Facial Expressions (AAAI 2024 · similarity 0.86)
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.
