Style Selector Node
738 A1111-style prompt presets, no find-and-replace node needed
- clip
- Positive CONDITIONING
- Negative CONDITIONING
- CLIP
- Positive text
- Negative text
If you've spent any time in Automatic1111 or Fooocus, you know the style dropdown: pick "Neonpunk" or "Analog film" from a list and it silently appends a wall of tuned keywords to your prompt. ComfyUI never got that natively - it's a node graph, not a settings panel - but this node brings the whole idea over, and it does it without the usual song and dance. Most ComfyUI ports of style-preset systems need you to load a JSON file and then wire up a separate find-and-replace node to inject your prompt into the {prompt} placeholder. This one skips that: type your prompt, pick a style, done.
How it works
Under the hood it's built on the A1111 Prompt Styler style library - the same lineage as twri's SDXL Prompt Styler, which the README credits directly. Each style is a template pair (a positive suffix and a negative suffix) that gets merged with whatever you typed. Pick "Fantasy art" and your prompt gets fantasy-art keywords stitched on; pick "None" and it passes through untouched. With 738 styles bundled, this is a genuinely huge library - way past what any single UI ships by default - so treat the dropdown as a starting point to skim, not something you'll memorize.
The clever bit is the optional CLIP input. Connect it and the node runs the styled text through CLIP text encoding for you, handing back ready-to-sample CONDITIONING. Leave it disconnected and you just get the styled text back as strings - useful if you want to inspect what a style actually does to your prompt, or route the text somewhere else first (another styler, a text-cleanup node, whatever).
The inputs and outputs that matter
Positive/Negative- your raw prompt text, same as any text-to-conditioning setup.styles- the dropdown. Defaults toNone. It's a flat list of 738 entries, not grouped by category, so know roughly what you're looking for before you open it (searching by typing usually works in ComfyUI's combo widgets).clip(optional) - plug in your checkpoint's CLIP if you want conditioning out directly.
Five outputs come off this node: Positive CONDITIONING and Negative CONDITIONING (only meaningful if CLIP was connected), a passthrough CLIP output so you don't have to split your CLIP wire elsewhere in the graph, and Positive text / Negative text - the actual styled strings, handy for a Show Text node if you want to see what the style added.
How to install it
Search KGnodes in ComfyUI Manager, or:
cd ComfyUI/custom_nodes
git clone https://github.com/shahkoorosh/ComfyUI-KGnodes.git
pip install -r requirements.txt
Restart, then find it under Add Node → 🎨KG. No models, no downloads - the style library ships as a bundled file inside the node pack.
Common issues & troubleshooting
Conditioning outputs are empty or the graph errors downstream. This almost always means CLIP isn't connected. The node is designed to degrade gracefully to text-only mode without CLIP, but if something downstream is expecting real conditioning, you'll need to wire the CLIP input in.
738 styles is a lot to scroll through. There's no category grouping in the node itself. If you already know roughly what you want ("anime," "cinematic," "isometric"), typing into the dropdown to filter is faster than scrolling. If you're just browsing, expect to spend a minute the first time finding your favorites.
A style barely changes the output. These presets are keyword suffixes, and how much a suffix moves the needle depends entirely on your checkpoint and how strongly you're weighting your own prompt. A style stacked onto a prompt that's already packed with its own detailed descriptors will have less room to push things around than a short, plain prompt will. This isn't specific to this node - it's how every prompt-styler-style tool behaves, going back to A1111's original.
Style text looks like it's fighting your own prompt. The styler appends rather than rewrites, so if your prompt and the style both make claims about mood or medium (you wrote "photorealistic" and picked "Comic book"), you'll get a tug-of-war. Pick a style that complements what you already wrote, or lean on the style alone and keep your own prompt to the subject.
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| Positive | STRING | — | |
| Negative | STRING | — | |
| styles | COMBO | None | 738 options: None, Enhance, Anime, Photographic, Digital art, Comic book, +732 |
| clipopt | CLIP | — |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| Positive CONDITIONING | CONDITIONING | — |
| Negative CONDITIONING | CONDITIONING | — |
| CLIP | CLIP | — |
| Positive text | STRING | — |
| Negative text | STRING | — |