NAI 提取线稿
Turn a finished render into a clean lineart, one call away
- image
- 图像
- 统计
NAI 提取线稿 (Line Art) is the simplest node in the Director Tools family and probably the most useful in a real comic pipeline: it takes an image and asks NovelAI to extract the line art from it - a clean, inkable line version of the subject. No prompt, no knobs, no settings. One IMAGE in, one IMAGE out, plus a 统计 string.
That's the whole input surface, which is exactly right. There's no strength, no style choice, no mode - NovelAI's lineart tool does one thing and this node gets out of its way. The output is the linework at roughly the source resolution, ready to feed into the colorize node, an inking workflow, or a re-render pipeline.
Where it earns its keep
The obvious loop is: render a full-color illustration → extract lineart → colorize it again with a different palette. It's also the cleanup step - if you want to produce an ink page from a colored render, or you need clean linework as an input for further editing, this is the one-click door. For the reverse direction (lineart → colored), the NAI 自动上色 node is the natural pair; the two of them are the front and back of the same comic-style round trip.
Honest limits: it's a one-shot API call with no dials, so there's no "extract more detail" or "cleaner lines" control - the only adjustments are the source image itself. A noisy, heavily textured source extracts noisier lines; a clean render extracts clean lines. And like every Director Tool it costs Anlas and only runs when you execute the node. For a genuinely difficult line-extraction job, the local preprocessor nodes in the broader ComfyUI ecosystem (the lineart preprocessors that feed ControlNet) give you far more control - but they run on your GPU and you get to tweak them. This node is the paid, zero-config version, and for "I just want the lineart now" it's hard to beat.
Install
Standard: search ComfyUI-NovelAI-GENYTOOLS in ComfyUI Manager, or git clone https://github.com/XTOGENY/ComfyUI-NovelAI-GENYTOOLS.git into custom_nodes + pip install -r requirements.txt, Persistent API Token in settings. Wire it between your render and the saver and you've got a two-node comic pipeline.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| 图像 | IMAGE | — |
| 统计 | STRING | — |