Research map / Video / Video generation
RIGID: Recurrent GAN Inversion and Editing of Real Face Videos
Yangyang Xu, Shengfeng He, Kwan-Yee K. Wong, Ping Luo
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 temporal compensated code from recurrent encoder (ablation) | RIGID · RAVDESS-72 video inversion | WE (warp error): 106.52 -> 84.62 | Improved |
| add learned noise map (ablation) | RIGID · RAVDESS-72 video inversion | MSE (x100): 2.28 -> 1.04 | Improved |
| add latent frequency disentanglement (share latter latent code across frames) (ablation) | RIGID · RAVDESS-72 video inversion | WE (warp error): 103.61 -> 84.62 | Improved |
| add temporal consistency loss L_tc (ablation) | RIGID · RAVDESS-72 video inversion | WE (warp error): 122.08 -> 84.62 | Improved |
| add in-between frame composition constraint (ablation) | RIGID · RAVDESS-72 video editing | FVD: 299.12 -> 198.54 | Improved |
| add in-between frame composition constraint (ablation) | RIGID · RAVDESS-72 video inversion | WE (warp error): 705.22 -> 84.62 | Improved |
| add in-between frame composition constraint on MSE (ablation) | RIGID · RAVDESS-72 video inversion | MSE (x100): 0.93 -> 1.04 | Hurt |
| use learned recurrent encoder instead of per-video optimization (STIT) | RIGID · video inversion inference time | IT (s): 851.5 -> 54.5 | Improved |
| RIGID versus optimization-based STIT/TCSVE on reconstruction | RIGID · RAVDESS-72 video inversion | MSE (x100): 1.04 vs 0.99 | Hurt |
Later papers that compared with this method
1 later accepted paper reports RIGID as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Training-free Detection of Generated Videos via Spatial-Temporal Likelihoods (CVPR 2026 · 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
- RIGID: Recurrent GAN Inversion and Editing of Real Face Videos and Beyond (IJCV 2025 · similarity 0.99)
- From Continuity to Editability: Inverting GANs With Consecutive Images (ICCV 2021 · similarity 0.92)
- High-Fidelity GAN Inversion for Image Attribute Editing (CVPR 2022 · similarity 0.90)
- In-N-Out: Faithful 3D GAN Inversion with Volumetric Decomposition for Face Editing (CVPR 2024 · similarity 0.90)
- VIVE3D: Viewpoint-Independent Video Editing Using 3D-Aware GANs (CVPR 2023 · similarity 0.89)
- Gradual Residuals Alignment: A Dual-Stream Framework for GAN Inversion and Image Attribute Editing (AAAI 2024 · similarity 0.89)
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.
