Research map / Video / Video generation
Scaling Zero-Shot Reference-to-Video Generation
Zijian Zhou, Shikun Liu, Haozhe Liu, Haonan Qiu, Zhaochong An, Weiming Ren, Zhiheng Liu, Xiaoke Huang, Kam-Woh Ng, Tian Xie, Xiao Han, Yuren Cong and 5 more
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 |
|---|---|---|---|
| train with masked video frames as references instead of OpenS2V-5M R2V data (ablation) | Saber · OpenS2V-Eval | Total Score: 56.24% -> 57.91% | Improved |
| use diverse mask shapes (ellipse, Fourier, polygon) instead of ellipse only (ablation) | Saber · OpenS2V-Eval | Total Score: 54.56% -> 57.91% | Improved |
| combine all mask types instead of Fourier only (ablation) | Saber · OpenS2V-Eval | Total Score: 56.33% -> 57.91% | Improved |
| combine all mask types instead of polygon only (ablation) | Saber · OpenS2V-Eval | Total Score: 56.49% -> 57.91% | Improved |
| sample foreground area ratio randomly instead of fixed r=0.3 (ablation) | Saber · OpenS2V-Eval | Total Score: 51.73% -> 57.91% | Improved |
| randomly sample foreground ratio versus fixed (ablation) | Saber · OpenS2V-Eval | NexusScore: 39.20% -> 47.22% | Improved |
| add spatial mask augmentation (rotation, scale, flip, shear) (ablation) | Saber · qualitative | copy-paste artifacts | Improved |
| add attention mask over reference tokens (ablation) | Saber · qualitative | gray artifacts around subjects | Improved |
| use zero-shot masked training instead of explicit R2V data methods (vs VACE-14B) | Saber · OpenS2V-Eval | Total Score: 57.55% -> 57.91% | Improved |
| zero-shot Saber versus Phantom-14B on subject consistency | Saber · OpenS2V-Eval | NexusScore: 37.43% -> 47.22% | Improved |
| scale number of reference images to about 12 | Saber · qualitative | composition quality | Hurt |
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.
- MAGREF — MAGREF: Masked Guidance for Any-Reference Video Generation with Subject Disentanglement (ICLR 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
- TI2V-Zero: Zero-Shot Image Conditioning for Text-to-Video Diffusion Models (CVPR 2024 · similarity 0.86)
- A Recipe for Scaling up Text-to-Video Generation with Text-free Videos (CVPR 2024 · similarity 0.86)
- OpenVid-1M: A Large-Scale High-Quality Dataset for Text-to-video Generation (ICLR 2025 · similarity 0.85)
- Movie Weaver: Tuning-Free Multi-Concept Video Personalization with Anchored Prompts (CVPR 2025 · similarity 0.85)
- ARGUS: Stacked Multi-View Identity Mosaic Injection for Subject-Preserving Video Generation (NeurIPS 2026 · similarity 0.85)
- FreeMask: Rethinking the Importance of Attention Masks for Zero-Shot Video Editing (AAAI 2025 · 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.
