InstantRetouch: Efficient and High-Fidelity Instruction-Guided Image Retouching with Bilateral Space
2026-06-03 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors developed a faster and more accurate way to adjust photo colors and tones using language instructions. Instead of changing the image directly, their method predicts a small grid of adjustments that can be applied to the whole image efficiently. They use knowledge from a complex diffusion model to guide these adjustments while keeping the photo looking natural and true to the original. Their approach avoids common problems like losing image details and works faster than recent similar methods. They also created a new test set to evaluate how well their method follows instructions, keeps details, and runs quickly.
language-guided photo retouchingdiffusion modelsbilateral spaceaffine transformsVariational Score Distillationprompt alignmentimage fidelityinstruction followingimage editing efficiencygenerative models
Authors
Jiarui Wu, Yujin Wang, Ruikang Li, Fan Zhang, Mingde Yao, Tianfan Xue
Abstract
Language-guided photo retouching aims to adjust color and tone while preserving geometry and texture. Recently, diffusion-based retouching shows a superior visual quality, but often struggles with both fidelity issues due to its generative nature and efficiency because of its iterative sampling process. In this work, we propose an efficient and fidelity-preserving retouching method using bilateral space manipulation, which is both compact and content-decoupled. Specifically, instead of directly editing pixels or image latents, our model predicts a low-resolution bilateral grid of affine transforms, which are sliced using a learned guidance map and then applied to the full-resolution image. This approach yields both high fidelity and improved efficiency. To retain strong priors of a pretrained generative model, we distill a multi-step diffusion model into our bilateral grid framework using Variational Score Distillation, complemented by a prompt alignment loss to guide instruction-following behavior. Additionally, we introduce a new benchmark and evaluate our method across multiple dimensions: fidelity, instruction following, and efficiency. Compared to the latest retouch methods, like Gemini-2.5-Flash (Nano-Banana), our method can avoid content drift, significantly improve latency, and generate visually pleasing edits, while maintaining a high level of fidelity. Project page: https://openimaginglab.github.io/InstantRetouch/.