Nodes/ComfyUI Inspire Pack/AnimeLineArt Preprocessor Provider (SEGS)
ComfyUI Node Runs on cloud

AnimeLineArt Preprocessor Provider (SEGS)

Line-art control for per-region detailing

By ltdrdata·Created 3 years ago·Updated 9 months ago· 805
AnimeLineArt Preprocessor Provider (SEGS)
    • SEGS_PREPROCESSOR

    This node lets an anime line-art ControlNet run on each region Impact Pack detects, instead of over the whole frame. It's a provider: it makes no image and does nothing standalone. You wire its output into Impact Pack's ApplyControlNet (SEGS) node, and during a detailer pass it extracts clean anime-style line art from each detected region and feeds it to a line-art ControlNet, so the redraw follows the original lines.

    The reason you'd reach for it: Impact Pack detailing crops a region and resamples it at full resolution, which fixes detail but can wander from the original drawing. A line-art ControlNet pins the outlines, so a detailer pass on an anime face or figure keeps the exact contours while cleaning up the interior. It's the anime counterpart to using a depth or FaceMesh control in the same slot - pick the preprocessor that matches what you're trying to hold constant.

    How it works

    The AnimeLineArt preprocessor traces an image into crisp anime-style line art - the kind of clean, sparse outline that anime/illustration line-art ControlNets were trained on (as opposed to the denser "realistic" lineart or standard Canny edges). Feed that to a lineart ControlNet during the resample and the model is constrained to the same linework.

    The "Provider (SEGS)" wrapper is Inspire's part. SEGS is Impact Pack's per-detection data type (bbox + mask + crop). A preprocessor provider tells the SEGS ControlNet node which preprocessor to run per region - here, the anime lineart one. The actual lineart implementation lives in Fannovel16's ControlNet Auxiliary Preprocessors; Inspire adapts it into the SEGS flow.

    The inputs and outputs that matter

    There are no inputs and no widgets. That's the design - this node is a small "which preprocessor" object, and all the real settings (ControlNet strength, detailer denoise, resolution) live on the nodes downstream.

    The single output is SEGS_PREPROCESSOR. It plugs into the preprocessor slot of Impact Pack's ApplyControlNet (SEGS), which also takes the SEGS, a lineart ControlNet model, and the image. This provider is only the "how to preprocess each region" half of that.

    How to install it

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

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

    then restart. It's a Dr.Lt.Data pack (also the author of ComfyUI-Manager and the Impact Pack this feeds).

    For it to do anything you need two more packs: the Impact Pack (SEGS + ApplyControlNet (SEGS)) and ControlNet Auxiliary Preprocessors (the actual lineart model) - the Inspire README lists these SEGS wrappers as depending on Fannovel16's aux nodes. And you'll want an anime lineart ControlNet downloaded for your base model.

    Common issues & troubleshooting

    No preprocessor / import error. ControlNet Auxiliary Preprocessors isn't installed or is failing to load. Install it via Manager and restart; the model downloads on first use.

    Lines are too busy or too sparse. The result depends on the source region's contrast and detail. AnimeLineArt aims for clean anime outlines; if your source is a realistic photo it'll look off, and a realistic-lineart or Canny preprocessor may suit it better. There are several lineart flavors in the aux pack - this one is specifically the anime style, so match it to anime/illustration content.

    The detail pass ignores the line art. Check the downstream lineart ControlNet is loaded and its strength isn't zero. This node only extracts the lines; the ControlNet is what constrains the redraw.

    Wrong architecture ControlNet. A lineart ControlNet trained for SD1.5 won't behave on SDXL and vice versa. Keep the control model matched to your base model, or the constraint just adds noise.

    CategoryInspirePack/SEGS/ControlNet

    Inputs (0)

    No inputs

    Outputs (1)

    NameTypeDescription
    SEGS_PREPROCESSORSEGS_PREPROCESSOR