Manga2Anime LineArt Preprocessor Provider (SEGS)

This is the anime-lineart cousin of the Canny provider, and it exists for one narrow-but-real job: you're refining a region of an anime or manga-style image with an Impact Pack detail pass, and you want that refinement to follow the linework of the original rather than reinvent it. Manga2Anime is a lineart preprocessor tuned for exactly that art style - cleaner and more forgiving than Canny's hard edges, better at picking up the confident inked strokes of anime art.

Like every other node in the SEGS/ControlNet folder of the Inspire Pack, it doesn't process anything by itself. It emits a preprocessing recipe that Impact Pack's Apply ControlNet (SEGS) runs later, once your detector has found the crops.

Why this instead of Canny or realistic lineart

ControlNet lineart comes in flavors, and the flavor matters. Canny gives you thin, literal edge maps - great for architecture, too brittle for hand-drawn art. Realistic lineart is thicker and more natural. The anime/manga variants are trained to extract the kind of clean outline you'd see in an inked panel, and Manga2Anime specifically is the one people reach for on 2D art where a Canny map would come out either scribbly or full of screentone noise. If your detail pass is on an illustrated character, this is usually the better structural guide.

How it fits the SEGS flow

Quick mental model of the whole thing, because this node is meaningless out of context. Impact Pack's detailing loop is: detect a region → crop it → resample at full resolution → paste back. Apply ControlNet (SEGS) lets you condition that resample on structure. But at the moment you build the graph, the crops don't exist yet, so you can't preprocess them ahead of time. So instead of an edge map, you pass a provider - a SEGS_PREPROCESSOR that carries "run Manga2Anime lineart on each crop." The preprocessing happens per-region at detail time.

The actual lineart extraction is implemented in Fannovel16's ControlNet Auxiliary Preprocessors; Inspire is the glue that lets it live in the SEGS pipeline.

The inputs and outputs that matter

There aren't any inputs. Per its schema, this node has no parameters at all - Manga2Anime has nothing to tune, so it's a zero-config provider. You just drop it in.

The single output is SEGS_PREPROCESSOR, which connects to the preprocessor slot on Impact Pack's Apply ControlNet (SEGS). That's the entire interface: no image in, no image out, one wire.

How to install it

ComfyUI Manager: search ComfyUI Inspire Pack, install, restart. Manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/ltdrdata/ComfyUI-Inspire-Pack

then restart. It's a Dr.Lt.Data pack, same author as ComfyUI-Manager and the Impact Pack, so it's a safe install.

Common issues & troubleshooting

Nothing happens when you add it. Expected - it produces no output image and has no standalone effect. It only does something wired into an Apply ControlNet (SEGS) node inside an Impact Pack detailing graph.

Red node / "not found" after restart. You're missing dependencies. This provider needs comfyui_controlnet_aux for the lineart model, and it's useless without the Impact Pack to consume it. Install all three through Manager.

You still need an anime lineart ControlNet checkpoint. The provider only tells the pipeline how to preprocess. The actual conditioning model gets loaded and fed into Apply ControlNet (SEGS) separately - and it has to match your base architecture (an SD1.5 or SDXL lineart ControlNet, whichever you're generating on).

It's not doing much on realistic images. Wrong tool. Manga2Anime is trained for 2D/inked art; on photographs it'll extract weak or noisy lines. Use the Canny or a depth provider for realistic subjects and save this one for illustration.