Paint-Anything: Unified Any-Color Control for Image Generation and Editing
Paint-Anything learns a shared hex-prompt interface for any-color image generation and editing on FLUX.2-4B, improving ACBench scores by 85.3% and 28.3%.
Paint-Anything enables specifying an object's target color as any 24-bit hex value for both image generation and editing through a shared hex-prompt interface trained with object-level color supervision. The authors build the Paint-500K dataset from real images via object grounding, perceptual color labeling, and editing-pair synthesis, supplemented with pure-color anchors used only at high-noise timesteps. They also introduce the Any Color Benchmark (ACBench) with T2I and Edit tracks; on FLUX.2-4B the method improves ACBench-T2I and ACBench-Edit by 85.3% and 28.3% over the base model and achieves the best average CompColor score.
- Shared hex-prompt interface covers both generation and editing tasks
- Paint-500K dataset built from real images with object-level color supervision
- Pure-color anchors applied only at high-noise timesteps to handle shadow ambiguity
- ACBench introduced to measure object-level hex color fidelity
Full article193 words · extracted from arxiv.org · click to collapse
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. We present Paint-Anything, which learns a shared hex-prompt interface for generation and editing through object-level color supervision. We develop a data pipeline that constructs Paint-500K from real images through object grounding, perceptual color labeling, and editing-pair synthesis. Since shadows make real-image labels only approximate colors, we complement this supervision with pure-color anchors whose pixels exactly match their paired hex values. These anchors are used only at high-noise timesteps, leaving low-noise training to natural images. We further introduce Any Color Benchmark (ACBench), comprising ACBench-T2I and ACBench-Edit, to measure object-level hex color fidelity across both tasks. On FLUX.2-4B, Paint-Anything improves ACBench-T2I and ACBench-Edit scores by 85.3% and 28.3%, respectively, relative to the base model, with ablations supporting the training recipe. It also achieves the highest average CompColor score among the compared methods.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.20816