Research map / Image generation / Fast diffusion sampling
Diffusion in Style
Martin Nicolas Everaert, Marco Bocchio, Sami Arpa, Sabine Süsstrunk, Radhakrishna Achanta
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 |
|---|---|---|---|
| adapt the initial latent noise distribution to the target style before fine-tuning Stable Diffusion | Stable Diffusion · Anime sketch, few-shot Pokemon and other styles | normalized FID (style matching) | Improved |
| fine-tune on 50 images for 1000 steps with style-adapted noise instead of 15k steps on all 833 images | Text-To-Pokemon comparison · Few-shot Pokemon | style matching and prompt alignment: 15k steps on 833 images -> 1000 steps on 50 images | Improved |
| adapted model versus LoRA fine-tuning on Pokemon | Diffusion in Style · Few-shot Pokemon | style matching (qualitative) | Improved |
| adapt the model instead of prompt engineering with style description | Stable Diffusion · Styles 3 to 5 (48 Famous Americans, Dali, pictograms) | normalized FID | Improved |
| adapt Stable Diffusion versus applying Informative Drawings image translation on top, at high guidance weights | Diffusion in Style · Anime sketch | CLIP score and normalized FID | Hurt |
| increase classifier-free guidance weight (ablation) | Diffusion in Style · DrawBench prompts, six styles | CLIP score versus normalized FID | Mixed |
| constant spatial mean and variance, 30 BLIP captions per image and 250 iterations for very few images | Diffusion in Style · Starry Night style with 3 images | qualitative style quality: 1000 -> 250 iterations | Improved |
Closest papers by meaning
- Smooth Diffusion: Crafting Smooth Latent Spaces in Diffusion Models (CVPR 2024 · similarity 0.90)
- Tuning Timestep-Distilled Diffusion Model Using Pairwise Sample Optimization (ICLR 2025 · similarity 0.88)
- Diffusion Model is Effectively Its Own Teacher (CVPR 2025 · similarity 0.88)
- Tracing the Roots: Leveraging Temporal Dynamics in Diffusion Trajectories for Origin Attribution (NeurIPS 2025 · similarity 0.88)
- Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer (CVPR 2024 · similarity 0.88)
- Multistep Distillation of Diffusion Models via Moment Matching (NeurIPS 2024 · 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.
