AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

StyleDrop: Text-to-Image Synthesis of Any Style

Kihyuk Sohn, Lu Jiang, Jarred Barber, Kimin Lee, Nataniel Ruiz, Dilip Krishnan, Huiwen Chang, Yuanzhen Li, Irfan Essa, Michael Rubinstein, Yuan Hao, Glenn Entis and 2 more

NeurIPS 2023 · Image generation › Text-to-image diffusion · approach: Personalized and region-controlled generation · 16 citations (OpenAlex)

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 changedWhereMeasuredEffect
style-tune Muse with StyleDrop (HF) instead of plain Muse promptingMuse · 6 styles, Parti promptsCLIP Style score: 0.556 -> 0.694Improved
style-tune Muse with StyleDrop (HF) instead of plain Muse promptingMuse · 6 styles, Parti promptsCLIP Text score: 0.323 -> 0.322No clear effect
StyleDrop on Muse instead of DreamBooth on ImagenStyleDrop · 6 styles, Parti promptsCLIP Style score: 0.644 -> 0.694Improved
StyleDrop (R1) instead of DreamBooth on ImagenStyleDrop · User studystyle preference (%): 9.7% -> 86.0%Improved
add iterative training with human feedback to round 1 (ablation)StyleDrop · 6 styles, Parti promptsCLIP Text score: 0.313 -> 0.322Improved
add iterative training with human feedback to round 1 (ablation)StyleDrop · 6 styles, Parti promptsCLIP Style score: 0.705 -> 0.694Hurt
use CLIP feedback instead of human feedback for iterative training (ablation)StyleDrop · 6 styles, Parti promptsCLIP Text score: 0.322 -> 0.329Improved
use human feedback instead of CLIP feedback for iterative training (ablation)StyleDrop · 6 styles, Parti promptsCLIP Style score: 0.673 -> 0.694Improved
select synthetic images by human feedback instead of random for iterative training (ablation)StyleDrop · 6 styles, Parti promptsCLIP Style score: 0.678 -> 0.694Improved
use iterative training with careful selection (CLIP/human) instead of random (ablation)StyleDrop · banana/bottle ablationCLIP Text score: 0.322 -> 0.339 (CLIP), 0.328 (HF)Improved
use descriptive style descriptor instead of rare token on Muse (ablation)StyleDrop · 6 styles, Parti promptsCLIP Style score: 0.654 -> 0.705Improved
use descriptive style prompt instead of rare token on Muse (ablation)StyleDrop · qualitative, melting golden 3dvisual style consistencyNo clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy