Research map / Image generation / Text-to-image diffusion
Language-Free Generative Editing from One Visual Example
Omar Elezabi, Eduard Zamfir, Zongwei Wu, Radu Timofte
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 |
|---|---|---|---|
| remove condition steering (generate instead of steer) (ablation) | VDC · SR / DeRain (one-shot) | FID: 41.41 -> 55.08 (SR); 87.12 -> 122.20 (DeRain) | Hurt |
| add data augmentation to one-shot example (ablation) | VDC · SR / DeRain (one-shot) | FID: 48.53 -> 41.41 (SR); 131.82 -> 87.12 (DeRain) | Improved |
| add pixel-space loss alongside latent loss (ablation) | VDC · SR / DeRain (one-shot) | FID: 46.93 -> 41.41 (SR); 93.79 -> 87.12 (DeRain) | Improved |
| add MLP condition generator (ablation) | VDC · SR / DeRain (one-shot) | FID: 44.31 -> 41.41 (SR); 106.87 -> 87.12 (DeRain) | Improved |
| initialize condition generator from text instead of no text (ablation) | VDC · SR / DeRain (one-shot) | FID: 41.41 -> 56.88 (SR); 87.12 -> 119.59 (DeRain) | Hurt |
| per-step condition instead of single not-conditioned generator (ablation) | VDC · SR / DeRain (one-shot) | FID: SR 41.74 -> 41.41; DeRain 89.82 -> 87.12 (LPIPS SR 0.3048 -> 0.2666) | Mixed |
| extend diffusion path from 10% to 20-30% (ablation) | VDC · SR / DeRain (one-shot) | FID: 41.41 -> 56.26 / 58.12 (SR) | Hurt |
| shorten diffusion path to 5% (ablation) | VDC · SR (one-shot) | FID: 41.41 -> 56.86 | Hurt |
| use eight visual examples (multi-shot) instead of one | VDC · six tasks | FID: DeRain 87.12 -> 69.52; SR 41.41 -> 45.89 | Mixed |
| add inversion correction to multi-shot | VDC · six tasks | FID: DeRain 69.52 -> 66.92; DeNoise 88.58 -> 82.57 | Improved |
| replace text prompts with visual-example condition (one-shot VDC vs OmniGen) | VDC · DeNoise | FID: 150.80 -> 89.51 | Improved |
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.
- OmniGen — OmniGen: Unified Image Generation (CVPR 2025) · better on 6 of 6 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
- Textualize Visual Prompt for Image Editing via Diffusion Bridge (AAAI 2025 · similarity 0.92)
- EditInfinity: Image Editing with Binary-Quantized Generative Models (NeurIPS 2025 · similarity 0.91)
- Localizing and Editing Knowledge In Text-to-Image Generative Models (ICLR 2024 · similarity 0.90)
- RelationAdapter: Learning and Transferring Visual Relation with Diffusion Transformers (NeurIPS 2025 · similarity 0.90)
- Inversion-Free Image Editing with Language-Guided Diffusion Models (CVPR 2024 · similarity 0.89)
- ACE: All-round Creator and Editor Following Instructions via Diffusion Transformer (ICLR 2025 · 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.
