Paint-Anything: Unified Any-Color Control for Image Generation and Editing via Hex Prompts

Forum topic · 小凯 · 2026-09-19

Summary

Paint-Anything is a new approach enabling precise any-color control in image generation and editing by allowing users to specify any 24-bit hex value as an object's target color. Built on the observation that even compact large language models can associate hex values with color semantics, the method learns a shared hex-prompt interface for both generation and editing through object-level color supervision. The authors construct Paint-500K, a dataset built from real images via object grounding, perceptual color labeling, and editing-pair synthesis. Because shadows make real-image color labels only approximate, they supplement supervision with solid-color anchors whose pixels exactly match paired hex values, using these anchors only at high-noise timesteps while reserving low-noise training for natural images. They also introduce ACBench (Any-Color Benchmark), comprising ACBench-T2I and ACBench-Edit, to measure object-level hex color fidelity in both tasks. On FLUX.2-4B, Paint-Anything improves ACBench-T2I and ACBench-Edit scores by 85.3% and 28.3% respectively, and achieves the highest average CompColor score among compared methods. Paper: arXiv 2609.20816.

Paper Overview

Field: Computer Vision (CV) Authors: Ji Xie, Dewei Zhou, Xinyu Huang, Zhennan Chen, Xun Wang Published: 2026-09-17 arXiv: 2609.20816

Abstract

Professional design requires any-color control: the ability to specify an object's target color with any 24-bit hex value for image generation and editing. Prior work has explored color generation, editing, and colorization, but often relies on dedicated color representations or specialized inference procedures. Advances in large language models offer a simpler starting point: even compact models can associate hex values with color semantics.

The authors present Paint-Anything, which learns a shared hex-prompt interface for generation and editing through object-level color supervision.

Key Contributions

Paint-500K dataset: Built from real images through a data pipeline involving object grounding, perceptual color labeling, and editing-pair synthesis.
Solid-color anchors: Since shadows make real-image labels only approximate colors, supervision is supplemented with solid-color anchors whose pixels exactly match paired hex values. These anchors are used only at high-noise timesteps, while low-noise training is left to natural images.
ACBench (Any-Color Benchmark): Introduces ACBench-T2I and ACBench-Edit to measure object-level hex color fidelity in both generation and editing tasks.

Results

On FLUX.2-4B, Paint-Anything improves ACBench-T2I and ACBench-Edit scores by 85.3% and 28.3%, respectively.
Ablation studies support the proposed training scheme.
It achieves the highest average CompColor score among all compared methods.

---

*Auto-collected on 2026-09-19*