Line_Art Styler
47 dropdown presets for pushing a prompt toward pure line work
- text_positive
- text_negative
Want an image that comes out looking like an ink drawing, a technical blueprint, or a Zentangle rather than a photo? That's this node's whole job. Line_ArtStyler from aegis72/comfyui-styles-all is a single-dropdown prompt styler: you type your subject, pick one of 47 line-art styles, and it rewrites your prompt into line-art language before it hits the text encoder.
What it actually is
It's one of the small per-theme stylers in the pack - a preserved clone of azazeal04's removed SDXL Prompt Styler styles. Dropdown options run from Architectural Blueprint and Botanical Illustration through Fashion Sketch, Intricate Mandala, Manga Character Sketch, Sci-Fi Spaceship Blueprint, and Vehicle Technical Drawing. Each one is a text template that wraps your prompt in style keywords and appends a matching negative. Pick Architectural Blueprint and your subject gets framed as a precise drafting-style drawing, with the opposite pushed into the negative prompt.
Text in, text out. No model, no API key, nothing to download. It does exactly one thing: it makes your prompt sound like line art.
The inputs that matter
text_positive- your subject. "a steam train," "a fox," whatever you're drawing. This is the text the style wraps.Line_Art- the 47-entry dropdown. The first option isNone, which passes your prompt through untouched - handy when you want the node in the graph but idle.log_prompt- defaults to Yes and prints before/after prompts to the console. Flip it to No once you trust the output.
The node returns two STRINGs: text_positive (your styled prompt) and text_negative (your negative plus the style's own negative, comma-joined). Wire them into whatever CLIPTextEncode your checkpoint uses.
A honest word about what it won't do
This biases the prompt toward line-art rendering; it doesn't force structural line art the way a ControlNet lineart model does. If you need precise contour control - say, turning a photo into a clean drawing with recognizable edges - you'll get more mileage pairing this prompt style with a ControlNet lineart/scribble unit than from the text alone. Use this node for flavor and vibe: a consistent "everything is ink on paper" look across a batch.
Installing it
ComfyUI Manager (search ComfyUI-styles-all) or:
cd ComfyUI/custom_nodes
git clone https://github.com/aegis72/comfyui-styles-all
Then restart ComfyUI. There's no requirements.txt and no model file - the whole pack is stdlib Python plus JSON.
Where people get burned
- It looks like it's "SDXL only," but it isn't. It just prepends/appends text, so it works on SD 1.5 models too. The one place it underperforms is Flux and other LLM-encoded models, which don't respond to this style of keyword salad.
- Dropdown won't show new options you added? Restart. The menu is built from the JSON files in
custom_nodes/comfyui-styles-all/data/Line_Art/at load time. - Negatives stacking up. The style appends its own negative to yours. That's usually good, but if you already have a long negative list, the combined string can get heavy - trim your own.
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| text_positive | STRING | — | |
| text_negative | STRING | — | |
| Line_Art | COMBO | 47 options: None, Abstract Geometric Shapes, Abstract Line Dance, Abstract Line Haiku, Abstract Line Meditation, Abstract Line Poetry, +41 | |
| log_prompt | BOOLEAN | true | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| text_positive | STRING | — |
| text_negative | STRING | — |