MagicColor
HKUST diffusion model — one-click auto-coloring of line art with a reference image
An HKUST diffusion model: one-click automatic coloring of line art with a reference image — for comics and illustration.
Image generation and computer vision — detection, editing, 3D.
Key features
- Diffusion-based framework for multi-instance sketch colorization with consistency
- Colors multiple instances at once from reference images and masks, faster than
- Pre-trained weights downloadable via download.sh; training supported on RTX 4090/A100
- Gradio interface for inference; automatic mask extraction tooling included
- Reference contributions from MangaNinjia, ColorizeDiffusion, DreamBooth
Good for
- Anime/manga artists colorizing line art with consistent
- Researchers training sketch colorization models
Stack & tags
PythonPyTorchsketch-colorizationanimediffusionimage-generationgradioopen-weightsmulti-instance
Quick start
bash download.sh
cd scripts
bash multi_ref_infer.shFrom the project README on GitHub.
Similar open-source projects
FLUX.2 klein
★ 25,996Black Forest Labs' sub-second image generation — generate and edit in <0.5s
DiffSynth-Studio
★ 13,200The framework behind Qwen-Image-i2L — the first "Image→LoRA" visual generation system