cs.CVOct 4, 2026

Salvation Lies Within: Eliciting Inherent Style Transfer in Step-Distilled Diffusion Models

Authors: Shengyin Sun, Yiming Li, Yingzhao Lian, Xing Li, Xingzhi Zhou, Anxin Tian, Zhili Wang, Haoyang Li, +2 more

Organizations: Huawei Technologies · Hong Kong Polytechnic University · City University of Hong Kong

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Adapting step-distilled text-to-image (T2I) models through post-training incurs additional computational costs and affects native few-step generation behavior. This motivates a complementary route beyond style-specific adaptation: drawing on the visual knowledge already encoded in step-distilled T2I models to elicit stylistic capabilities through language. Pursuing this direction requires textual guidance that captures how visual attributes jointly define a style and remain applicable as the depicted content changes. To explore this approach, we introduce StyleForge, a fully automatic, training-free framework that expresses reference styles as reusable rendering instructions. By integrating overall rendering characteristics with local color and lighting behavior, StyleForge organizes visual evidence from reference images into a coherent specification of how the target style should be expressed. The specification is then compiled into textual guidance that can be reused across content prompts, enabling frozen step-distilled T2I models to render different subjects and scenes in the reference style while retaining native few-step generation. Extensive experiments show relative gains of up to 29.47% in generation quality scores over the strongest baseline, while Pareto analysis indicates that improved stylization is accompanied by strong adherence to the requested content.

Figures & tables

Figure 1: Precise textual control of rendering attributes.
Figure 2: Overview of StyleForge. (1) The agent extracts structured style evidence from reference images. (2) Color and luminance statistics ground the extracted evidence. (3) The agent compiles a reusable style prompt and combines it with new content prompts to guide a step-distilled T2I model.
CategoryDescriptionTransfer Policy
Rendering MRM_{\mathcal{R}}Medium, Surface, Line, Shape, Texture, Detail Density, etc.Preserve ✓
Color and Light ARA_{\mathcal{R}}Carrier Surfaces, Extent, Distribution of Colors, Light, etc.Preserve ✓
Composition TRT_{\mathcal{R}}Layout, Depth, Negative Space, etc.Weaken ↓\bm{\downarrow}
Reference Content LRL_{\mathcal{R}}Subjects, Props, Scenes, Enclosing Structures, etc.Exclude ✗
Table 1: Transfer policies for the four types of information in the style state SRS_{\mathcal{R}} .
Figure 3: Qualitative comparison with baselines on FLUX.2-Klein-4B.
Figure 4: Qualitative comparison with baselines on FLUX.2-Klein-9B.
MethodQwen3-VL-8B-InstructQwen3-VL-32B-InstructQwen3.5-9BAvg. Rank
ScoreWin Rate (%)ScoreWin Rate (%)ScoreWin Rate (%)
FLUX.2-Klein-4B Backbone
DSE3.22652.67%3.18746.75%2.07746.06%3.667
RCG2.68941.84%2.95047.53%1.69844.12%4.500
Vanilla SFT3.76940.54%3.30029.43%1.50634.86%4.167
PSO3.75737.82%3.25527.49%1.49632.65%5.167
Table 2: VLM evaluation of content and style transfer on FLUX.2-Klein backbones ( Best , Second Best ). Gain is relative to the strongest baseline.
Figure 5: Pareto comparison of content and style across FLUX.2-Klein backbones.
Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 6: Additional qualitative comparison with baselines on FLUX.2-Klein-4B.
Figure 7: Additional qualitative comparison with baselines on FLUX.2-Klein-9B.
Figure 8: Supplementary qualitative comparison with i2L on the FLUX.2-Klein-4B series. Each group shows style references, StyleForge using the step-distilled model, and i2L using the base model with CFG scales of 1 and 4.
Figure 9: Qualitative comparison with baselines on Z-Image-Turbo 6B.
Figure 10: Qualitative style-transfer results with GPT-Image-2.5-Flare. The first column shows style references, and the remaining columns show generated images for different content prompts.
Figure 12: Qualitative illustration of subject driven generation with a step-distilled model. Three reference subjects, a can, a toy car, and a berry bowl, are placed in a new cinematic scene. The reference-conditioned result preserves their colors, graphics, shapes, and lettering, while the text-only sample uses the same prompt and seed without access to the reference images.
Figure 13: Additional Pareto comparisons using Qwen3-VL-8B-Instruct and Qwen3.5-9B as judges. The two left panels show the FLUX.2-Klein-4B and 9B results for Qwen3-VL-8B-Instruct, while the two right panels show the corresponding results for Qwen3.5-9B.

Explore similar work

CardsList

  1. STEPS: Scene Text Editing with Preserved Style Using Diffusion and Contrastive Style Encoding

    Sep 29, 2026Nicolas Thiebaut, Nameer Hirschkind, Xiao Yu +1Text-To-Image Diffusion ModelsStyle

    Roblox San Mateo, CA, USA

  2. RL-Native Distillation: Exploiting Scored Trajectories for Few-Step Image Generation

    Aug 10, 2026Yuhan Li, Fangao Zeng, Sicong Kang +7Diffusion-Based Reinforcement Learning MethodsImage Generation

    Shanghai Jiao Tong University · Taobao & Tmall group of Alibaba

  3. Qwen-Image-Flash: Rethinking the Training Recipe for Few-Step Distillation

    Jun 2, 2026Tianhe Wu, Zikai Zhou, Kun Yan +23Image EditingGenerative Models