🪄 Style Selector (OreX)

Remember how Fooocus ships with a dropdown of named styles - "Fooocus Cinematic", "sai-3d-model" - and you just pick one instead of hand-typing a wall of style tokens? OrexStyleSelector brings that whole idea into ComfyUI, with a proper UI: a thumbnail grid of styles you click to select, complete with a search bar and a favorites star. Pick a style, and out comes a complete positive and negative prompt you wire straight into a CLIP Text Encode node. If you've ever gotten tired of re-pasting the same style blobs into every prompt, this is the node that kills that chore.

It's part of the comfyui-OreX pack (by the Russian devs OreX and Pavel K.), which is a broad grab-bag of load/save, LLM, and image tools. The pack's README only documents the load/save nodes, so don't go looking for this one there - the code is ahead of the docs.

How it works

The node is a thin but well-made wrapper around JSON files bundled in the pack's styles/ folder. Each entry is {name, prompt, negative_prompt, thumbnail}. Pick one and the node merges it with your own prompt. The {prompt} placeholder inside a style's prompt gets replaced with whatever you feed into the optional positive input; the style's negative_prompt is appended to your negative. No network calls, no models - it's string assembly plus a frontend.

The bundled libraries are the real selling point:

• fooocus_styles - 275 entries, the actual Fooocus style set (Enhance, Sharp, Masterpiece, sai-3d-model, analog film…).
• krea2_style and krea2_pose - Krea-flavored looks plus 120 pose directives.
• emotions - 7 emotional-expression prompts (Happiness, Anger, Surprise…).
• Klein_Edit - 16 edit-direction prompts ("Make Anime", "White background", "Remove People") with a nifty before/after slider preview.
• favorite - your own five-starred picks; favorite.json is created on first launch if it's missing.

Inputs and outputs that matter

• styles: which library to browse (defaults to favorite).
• batch_mode (default off): the important one. Off = one merged prompt, all selected styles comma-joined. On = a list with one prompt per selected style - feed that into a batch save loop and you get "one image per style" grids without touching a batch node.
• select_styles: comma-separated style names, or a JSON array. Leave it empty and the node just passes your prompt through unchanged - which is a safe default but trips people up ("why is nothing happening?").
• preview_scale: thumbnail size in the embedded grid, 0.5–2.
• Optional positive / negative inputs: your base prompt.

All three outputs - positive, negative, and file_name - are lists. file_name is derived from each style's thumbnail filename, so in batch mode you get a per-style name you can use in a Save Image node's filename field. That pairing (this node → orex Save Image) is clearly the intended workflow.

Install

Via ComfyUI Manager (search "comfyui-OreX"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/orex2121/comfyui-OreX

Then restart ComfyUI. Dependencies are Pillow and soundfile (soundfile is for the pack's audio node; this one needs nothing extra). No model downloads - the styles ship inside the repo.

Common issues

• Nothing changes when you run it. select_styles is empty, so it's a pass-through. Select something in the grid first.
• Typo'd style names vanish silently. If you hand-type a name that isn't in the JSON, the node skips it with no error. In batch mode you just get fewer outputs.
• Outputs are lists, always. Most string nodes handle them fine, but a naive string-concat node may choke. Keep positive/negative wired into a CLIP Text Encode.
• Stale install weirdness. The README's first line says delete the old comfyui-OreX folder and reinstall when upgrading - do that before reporting bugs.