Research map / Video / Video generation
MagicMirror: ID-Preserved Video Generation in Video Diffusion Transformers
Yuechen Zhang, Yaoyang Liu, Bin Xia, Bohao Peng, Zexin Yan, Eric Lo, Jiaya 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 |
|---|---|---|---|
| add Conditioned Adaptive Normalization (ablation) | MagicMirror · ablation (half training) | ID similarity: 0.886 -> 0.911 | Improved |
| add Conditioned Adaptive Normalization (ablation) | MagicMirror · ablation (half training) | FM_inter: 0.568 -> 0.665 | Improved |
| add image pre-training (ablation) | MagicMirror · ablation (half training) | ID similarity: 0.883 -> 0.911 | Improved |
| add image pre-training (ablation) | MagicMirror · ablation (half training) | FM_inter: 0.559 -> 0.665 | Improved |
| use dual-branch (identity + structural) face embedding instead of identity branch only (ablation) | MagicMirror · ablation (half training) | FM_inter: 0.572 -> 0.654 | Improved |
| use dual-branch instead of structural branch only (ablation) | MagicMirror · ablation (half training) | FM_inter: 0.584 -> 0.654 | Improved |
| use dual-branch instead of single branch (ablation) | MagicMirror · ablation (half training) | ID similarity: 0.865 -> 0.870 | Improved |
| add masks and CAN pipeline on top of dual branch to reach full model (ablation) | MagicMirror · ablation (half training) | ID similarity: 0.870 -> 0.911 | Improved |
| evaluate ID-Animator at 512x512 instead of 480x720 | ID-Animator · ID-preserved video generation | ID similarity: 0.923 -> 0.921 | No clear effect |
| MagicMirror vs CogVideoX-I2V | CogVideoX-I2V · video generation | Dynamic Degree: 0.660 -> 0.705 | Improved |
| MagicMirror vs ConsisID | ConsisID · video generation | FM_ref: 0.652 -> 0.730 | Improved |
| MagicMirror vs ID-Animator on ID similarity | ID-Animator · video generation | Average ID Similarity: 0.923 -> 0.922 | No clear effect |
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.
- DynamiCrafter — DynamiCrafter: Animating Open-domain Images with Video Diffusion Priors (ECCV 2024) · better on 2 of 2 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
- StableAnimator: High-Quality Identity-Preserving Human Image Animation (CVPR 2025 · similarity 0.89)
- Identity-Preserving Text-to-Video Generation by Frequency Decomposition (CVPR 2025 · similarity 0.89)
- FaithfulFaces: Pose-Faithful Facial Identity Preservation for Text-to-Video Generation (NeurIPS 2026 · similarity 0.87)
- Frame In-N-Out: Unbounded Controllable Image-to-Video Generation (NeurIPS 2025 · similarity 0.87)
- VideoMage: Multi-Subject and Motion Customization of Text-to-Video Diffusion Models (CVPR 2025 · similarity 0.87)
- ARGUS: Stacked Multi-View Identity Mosaic Injection for Subject-Preserving Video Generation (NeurIPS 2026 · similarity 0.87)
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.
