Staying True to the Origin: Continuous Image Stylization with Smooth Transitions

2026-08-08

AI Summary

Existing image style transfer methods struggle to simultaneously preserve content structure, achieve strong stylization, and enable continuous control over style intensity. This work proposes a novel approach based on the Diffusion Transformer (DiT), employing a two-stage training strategy: first freezing the pretrained DiT backbone, then learning an anchor projector in a low-rank space. For the first time, spline interpolation is introduced to enable smooth, arbitrary style intensity transitions across multiple reference images. Requiring only discrete training samples, the method supports continuous, controllable, and visually consistent style transfer, significantly outperforming current state-of-the-art techniques while maintaining high content fidelity.

Abstract

Recent advances in generative models have achieved remarkable performance in text- and image-conditioned editing. However, preserving the content of a given image while referencing style patterns from another remains challenging, often leading to uncontrollable stylization results. In this paper, we approach image stylization from the perspective of continuous control, aiming to enable modern Diffusion Transformer (DiT)-based multi-reference editing models to (1) faithfully preserve the semantic structure of the content image, (2) render strong stylization effects, and (3) smoothly transition between the two. To this end, we propose a simple yet effective two-stage training strategy along with a style-strength-aware spline formulation. Specifically, in the first stage, the model is trained to produce strongly stylized outputs while preserving the content semantics as much as possible. In the second stage, with the base model frozen, we learn a set of anchor projectors that map various stylization strengths into the model parameter space. During inference, by performing style-strength-aware spline interpolation in a low-rank space, our method enables continuous control over stylization strength, even though the model is trained with only a few discrete strength levels. Extensive experiments demonstrate that our method supports precise and continuous manipulation of stylization strength while generating high-fidelity results with modern DiT models. Project page: https://reychiaro.github.io/StyleController.

Problem

• image stylization
• content preservation
• continuous control
• style transition
• semantic structure

Innovation

• continuous stylization
• Diffusion Transformer
• style-strength control
• spline interpolation
• two-stage training