Line_Art Styler (Advanced)

The same 47 line-art presets as the plain Line_Art Styler - Architectural Blueprint, Intricate Mandala, Vehicle Technical Drawing - but with two positive boxes and six outputs. Line_ArtStylerAdvanced from aegis72/comfyui-styles-all targets SDXL workflows that want a separate global prompt for the OpenCLIP-G encoder and a local prompt for CLIP-L.

The inputs

• text_positive_g / text_positive_l - your two positive boxes.
• text_negative - your negative; the style's is appended.
• Line_Art - the 47-option dropdown.
• negative_prompt_to - Both (default), G only, or L only.
• log_prompt - on by default; prints the before/after to the console.

Outputs: text_positive_g, text_positive_l, text_positive (combined), text_negative_g, text_negative_l, text_negative.

The catch with these templates

As with the pack's other Advanced nodes, two gotchas verified by running the code. The line-art templates are single-part wrappers - no . separator - so the G/L split has nothing to work with. And the two positive boxes trade places: your text_positive_l gets styled and exits as text_positive_g, while text_positive_g passes through unstyled to text_positive_l.

Path of least resistance:

• Put your subject in text_positive_l and use the combined text_positive output into CLIPTextEncode (or CLIPTextEncodeSDXL on SDXL).
• If you want one box and one dropdown, the plain Line_Art Styler avoids the swap entirely.

One more honest note: line-art prompt styles like these bias the render toward ink-and-paper output, but they won't guarantee clean contours. For real line-art structure, pair this with a ControlNet lineart unit - the prompt gives you the vibe, ControlNet gives you the lines.

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. No dependencies, no models.

Where people get burned

• The G/L swap - a styled output with the subject missing is the classic symptom.
• Restart after editing data/Line_Art/Line_Art.json for new options.
• SDXL-era keywords - weak on Flux and LLM-encoded models, same as the rest of the pack.