Research map / Image generation / Text-to-image diffusion
InstaStyle: Inversion Noise of a Stylized Image is Secretly a Style Adviser
Xing Cui, Zekun Li, Peipei Li, Huaibo Huang, Xuannan Liu, Zhaofeng He
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 |
|---|---|---|---|
| replace the inversion style noise with random noise (ablation) | InstaStyle · stylized generation, 100 images per style | style consistency score | Hurt |
| add prompt refinement (learned style token) on top of inversion noise (ablation) | InstaStyle · stylized generation, 100 images per style | style and content scores | Improved |
| use DDIM inversion noise as initial noise instead of random noise for generation (ablation) | Stable Diffusion · stylized generation | style details | Improved |
| use InstaStyle instead of Custom Diffusion | stylized text-to-image generation · 60 styles, 100 CIFAR100 objects | style consistency (CLIP): 0.611 -> 0.655 | Improved |
| use InstaStyle instead of StyleDrop | stylized text-to-image generation · 60 styles, 100 CIFAR100 objects | content fidelity (CLIP): 0.248 -> 0.294 | Improved |
| use InstaStyle instead of DreamBooth | stylized text-to-image generation · 60 styles, 100 CIFAR100 objects | content fidelity (CLIP): 0.244 -> 0.294 | Improved |
| use InstaStyle (text content) instead of style transfer methods that take content images | stylized generation · 60 styles | style consistency (CLIP): 0.604 -> 0.655 vs AesPA-Net | Improved |
| increase noise mix ratio and prompt mix ratio for style combination (ablation) | InstaStyle style combination · two-style combination | qualitative style | Improved |
Later papers that compared with this method
1 later accepted paper reports InstaStyle as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Wasserstein Style Distribution Analysis and Transform for Stylized Image Generation (ICCV 2025 · better on 4 of 4 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
- Style Aligned Image Generation via Shared Attention (CVPR 2024 · similarity 0.87)
- InitNO: Boosting Text-to-Image Diffusion Models via Initial Noise Optimization (CVPR 2024 · similarity 0.87)
- InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation (ICLR 2024 · similarity 0.87)
- Z*: Zero-shot Style Transfer via Attention Reweighting (CVPR 2024 · similarity 0.87)
- StyleSSP: Sampling StartPoint Enhancement for Training-free Diffusion-based Method for Style Transfer (CVPR 2025 · similarity 0.87)
- Conditional Balance: Improving Multi-Conditioning Trade-Offs in Image Generation (CVPR 2025 · similarity 0.87)
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.
