Research map / Video / Video generation
CanonSwap: High-Fidelity and Consistent Video Face Swapping via Canonical Space Modulation
Xiangyang Luo, Ye Zhu, Yunfei Liu, Lijian Lin, Cong Wan, Zijian Cai, Yu Li, Shao-Lun Huang
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 canonical-space warping before swapping (ablation) | CanonSwap · VFS benchmark | Pose error: 0.0227 -> 0.0159 | Improved |
| add canonical-space warping before swapping (ablation) | CanonSwap · VFS benchmark | Expression error: 0.9512 -> 0.7592 | Improved |
| add the refinement module before warping back (ablation) | CanonSwap · VFS benchmark | FVD: 481.77 -> 125.30 | Improved |
| add the refinement module before warping back (ablation) | CanonSwap · VFS benchmark | ID similarity: 0.4778 -> 0.5748 | Improved |
| add the soft spatial mask (Partial Identity Modulation) instead of global modulation (ablation) | CanonSwap · VFS benchmark | FVD: 165.17 -> 125.30 | Improved |
| add the soft spatial mask (ablation) | CanonSwap · VFS benchmark | ID retrieval: 97.38 -> 99.78 | Improved |
| add canonical warping (ablation) | CanonSwap · VFS benchmark | FVD: 131.57 -> 125.30 | Improved |
| use CanonSwap instead of strongest GAN baseline SimSwap | Video face swapping · VFS benchmark | LSE-C: 5.306 -> 6.053 | Improved |
| use CanonSwap instead of FSGAN on expression | Video face swapping · FF++ | Expression error: 0.7184 -> 0.7328 | No clear effect |
| frame-by-frame canonical swap instead of video diffusion | Video face swapping · Inference speed | FPS: 0.21 -> 14 (vs REFace) | Improved |
Later papers that compared with this method
1 later accepted paper reports CanonSwap as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- OmniFace: Bridging the Image-to-Video Gap for High-Fidelity Face Swapping via Diffusion Transformer (ECCV 2026 · better on 3 of 5 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
- Preserving Source Video Realism: High-Fidelity Face Swapping for Cinematic Quality (CVPR 2026 · similarity 0.92)
- DynamicFace: High-Quality and Consistent Face Swapping for Image and Video using Composable 3D Facial Priors (ICCV 2025 · similarity 0.92)
- VividFace: A Robost and High-Fidelity Video Face Swapping Framework (NeurIPS 2025 · similarity 0.92)
- OmniFace: Bridging the Image-to-Video Gap for High-Fidelity Face Swapping via Diffusion Transformer (ECCV 2026 · similarity 0.91)
- FaithfulFaces: Pose-Faithful Facial Identity Preservation for Text-to-Video Generation (NeurIPS 2026 · similarity 0.87)
- Synergizing Motion and Appearance: Multi-Scale Compensatory Codebooks for Talking Head Video Generation (CVPR 2025 · 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.
