Research map / Video / Video generation
PropFly: Learning to Propagate via On-the-Fly Supervision from Pre-trained Video Diffusion Models
Wonyong Seo, Jaeho Moon, Jaehyup Lee, Soo Ye Kim, Munchurl Kim
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 |
|---|---|---|---|
| use one-step clean latent estimation instead of full sampling for pair generation (ablation) | PropFly-1.3B · EditVerseBench-Appearance ablation | Pick score: 19.75 -> 20.35 | Improved |
| use one-step clean latent estimation instead of full sampling (ablation) | PropFly-1.3B · EditVerseBench-Appearance ablation | Video text alignment (ViCLIP): 24.77 -> 25.37 | Improved |
| use GMFM loss instead of standard FM loss (ablation) | PropFly-1.3B · EditVerseBench-Appearance ablation | Pick score: 19.50 -> 20.35 | Improved |
| use GMFM loss instead of standard FM loss (ablation) | PropFly-1.3B · EditVerseBench-Appearance ablation | Video text alignment (ViCLIP): 21.98 -> 25.37 | Improved |
| add Random Style Prompt Fusion (RSPF) (ablation) | PropFly-1.3B · EditVerseBench-Appearance ablation | Pick score: 20.28 -> 20.35 | Hurt |
| add Random Style Prompt Fusion (RSPF) (ablation) | PropFly-1.3B · EditVerseBench-Appearance ablation | Video text alignment (ViCLIP): 25.61 -> 25.37 | Hurt |
| train on on-the-fly pairs instead of Senorita-2M paired dataset (ablation) | PropFly-1.3B · EditVerseBench-Appearance ablation | Pick score: 19.53 -> 20.35 | Improved |
| train on on-the-fly pairs instead of paired dataset (ablation) | PropFly-1.3B · EditVerseBench-Appearance ablation | temporal consistency DINO: 97.85 -> 98.63 | Improved |
| raise high CFG scale from 2 to 7 in pair generation (ablation) | PropFly-1.3B · EditVerseBench-Appearance, high-CFG scale sweep | Pick score: 19.98 -> 20.35 | Improved |
| raise high CFG scale from 7 to 20 in pair generation (ablation) | PropFly-1.3B · EditVerseBench-Appearance, high-CFG scale sweep | Pick score: 20.35 -> 20.19 | Hurt |
| scale backbone from Wan-1.3B to Wan-14B (ablation) | PropFly · EditVerseBench-Appearance | Video text alignment (ViCLIP): 25.37 -> 26.05 | Improved |
| swap backbone from Wan-1.3B to LTX-2B (ablation) | PropFly · EditVerseBench-Appearance | Pick score: 20.35 -> 19.87 | 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.
- EditVerse — EditVerse: Unifying Image and Video Editing and Generation with In-Context Learning (ICLR 2026) · 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
- RFDM: Residual Flow Diffusion Models for Video Editing (CVPR 2026 · similarity 0.88)
- FADE: Frequency-Aware Diffusion Model Factorization for Video Editing (CVPR 2025 · similarity 0.87)
- VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling (NeurIPS 2026 · similarity 0.86)
- TokenFlow: Consistent Diffusion Features for Consistent Video Editing (ICLR 2024 · similarity 0.86)
- COVE: Unleashing the Diffusion Feature Correspondence for Consistent Video Editing (NeurIPS 2024 · similarity 0.85)
- 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.
