MegaStyle: Constructing Diverse and Scalable Style Dataset via Consistent Text-to-Image Style Mapping

Junyao Gao1,2*, Sibo Liu2*, Jiaxing Li3, Yanan Sun4, Yuanpeng Tu5,
Fei Shen6, Weidong Zhang2, Cairong Zhao1‡, Jun Zhang2‡
1Tongji University 2Tencent
3Nanyang Technological University 4Hong Kong University of Science and Technology
5University of Hong Kong 6National University of Singapore

*Equal contribution, ‡Corresponding Authors.

Visualizations of our style dataset (a)MegaStyle-1.4M and the stylized results produced by our style transfer model (b)MegaStyle-FLUX. MegaStyle-1.4M contains style pairs that share the style but have different content (intra-style consistency), as well as a large number of diverse styles (inter-style diversity). Trained on MegaStyle-1.4M, MegaStyle-FLUX effectively captures nuances—such as color, light, texture and brushwork—across various styles.

Paper (coming soon) Code (coming soon) Dataset (coming soon) Models (coming soon)

Abstract

In this paper, we introduce MegaStyle, a novel and scalable data curation pipeline that constructs an intra-style consistent, inter-style diverse and high-quality style dataset. We achieve this by leveraging the consistent text-to-image style mapping capability of current large generative models, which can generate images in the same style from a given style description. Building on this foundation, we curate a diverse and balanced prompt gallery with 170K style prompts and 400K content prompts, and generate a large-scale style dataset MegaStyle-1.4M via content–style prompt combinations. With MegaStyle-1.4M, we propose style-supervised contrastive learning to fine-tune a style encoder MegaStyle-Encoder for extracting expressive, style-specific representations, and we also train a FLUX-based style transfer model MegaStyle-FLUX. Extensive experiments demonstrate the importance of maintaining intra-style consistency, inter-style diversity and high-quality for style dataset, as well as the effectiveness of the proposed MegaStyle-1.4M.Moreover, when trained on MegaStyle-1.4M, MegaStyle-Encoder and MegaStyle-FLUX provide reliable style similarity measurement and generalizable style transfer, making a significant contribution to the style transfer community.

Data Curation Pipeline

Overview of our data curation pipeline. We first collect style and content images from open-source datasets. Next, we apply carefully designed instructions to generate style and content prompts with Qwen3-VL, together with balance sampling. Finally, we use Qwen-Image to generate style images using content-style prompt combinations. Please note that we use simplified content and style prompts for illustrative purposes only.

Visualizations of MegaStyle-1.4M

Each row presents the same style across different contents. MegaStyle-1.4M contains diverse, high-quality style images with strong intra-style consistency.

Stylized results of MegaStyle-FLUX

Trained on MegaStyle-1.4M, MegaStyle-FLUX generates stylized images that align with the content specified by the text prompt and the style of the reference image.

Comparisons

We compare MegaStyle-FLUX with the SOTA style transfer methods, including DEADiff, StyleShot, Attention-Distillation (Attn-Distill), CSGO, StyleCrafter, InstantStyle and StyleAligned. MegaStyle-FLUX achieves the superior performance compared to these baseline methods.

BibTeX

@article{
      TBA
    }