What is Line Art Coloring?

It's the task of adding color to line art.

If you can already draw your own line art, "how do I get AI to color my line art?" often matters more than text2image from scratch.


Line Art Coloring with ControlNet

With ControlNet Canny (edge), you can add color while keeping the input line art mostly intact.

flux_controlnet_union.json

Pass the line art (or an edge-detection result) to ControlNet, and describe the clothes, colors, and mood in the text prompt.


Line Art Coloring with Instruction-Based Image Editing

You can also hand line art straight to an instruction-based image editing model and have it do the coloring.

qwen_image_edit_2509.json

Just use the line art as the input image and give a text instruction like "color this line art in full color" or "color it in anime cel shading."


Reference-Based Coloring (Line Art + Color Image)

When you want to "color this line art with the same palette as this character" or "borrow the colors from this illustration," you can use reference-based coloring.

ControlNet (lineart / anime, etc.) locks in the shape and composition of the line art, while a reference image adapter like IP-Adapter takes a color illustration or photo and pulls the palette, texture, and art style toward it.

Some instruction-based image editing models that support multiple references also let you ask for things like "color the line art in image 1 using image 2 as reference."


Aside

Real artists work in a few more stages, like "base colors → shading → finishing."

And this goes beyond coloring: AI is good at producing a finished piece in one go, but not so good at building up a picture gradually the way humans do.

I'm Japanese, so I'm a little biased, but this kind of research and development is very active in Japan, from commercial services like Copainter to community LoRAs.