Research map / Video / Video generation
Zero-to-Hero: Empowering Video Appearance Transfer with Zero-Shot Initialization and Holistic Restoration
Tongtong Su, Chengyu Wang, Haipeng Liao, Jun Huang, Dongming Lu
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 Zero-Stage initialization as condition (zero-ref) instead of direct tgt-ref (ablation) | Hero-Stage · Blender-Color-Edit | avg PSNR: 19.38 -> 24.09 | Improved |
| add auxiliary target-frame pairs as implicit learning to zero-ref (ablation) | Hero-Stage · Blender-Color-Edit | avg PSNR: 24.09 -> 24.54 | Improved |
| use the target frame as explicit condition instead of implicit pairs (ablation) | Hero-Stage · Blender-Color-Edit | avg PSNR: 24.54 -> 26.76 | Improved |
| use the target frame as explicit condition instead of zero-ref only (1000 steps) (ablation) | Hero-Stage · Blender-Color-Edit | anchor-to-last-frame PSNR gap: -10.88 -> -4.32 | Improved |
| train directly anchor-to-reference (tgt-ref) for more steps (ablation) | Hero-Stage · Blender-Color-Edit | last target frame PSNR: 18.62 | No clear effect |
| use Mask-Attn instead of Pixel-Swap Zero-Stage initialization at t=200 (ablation) | Zero-to-Hero · Blender-Color-Edit | PSNR: 22.4982 -> 23.7836 | Improved |
| use Latent-Swap instead of Mask-Attn initialization at t=400 (ablation) | Zero-to-Hero · Blender-Color-Edit | PSNR: 26.7640 -> 26.9218 | Mixed |
| replace baselines with Zero-to-Hero on colorization | Zero-to-Hero · Colorization | PSNR: 23.4085 -> 28.2063 | Improved |
| replace CiA (β*) with Zero-to-Hero on General-Edit subject consistency | Zero-to-Hero · General-Edit | Subject Consistency: 0.9158 -> 0.8978 | Hurt |
Closest papers by meaning
- Zero-shot Image Editing with Reference Imitation (NeurIPS 2024 · similarity 0.89)
- VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling (NeurIPS 2026 · similarity 0.88)
- Zero-to-Hero: Enhancing Zero-Shot Novel View Synthesis via Attention Map Filtering (NeurIPS 2024 · similarity 0.87)
- Ground-A-Video: Zero-shot Grounded Video Editing using Text-to-image Diffusion Models (ICLR 2024 · similarity 0.86)
- FRESCO: Spatial-Temporal Correspondence for Zero-Shot Video Translation (CVPR 2024 · similarity 0.86)
- Transforming Image Editors into Video Editors (NeurIPS 2026 · similarity 0.85)
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.
