Milehigh Styler

MilehighStyler is the pack's namesake node and its big catch-all menu - about 1,500 styles in one dropdown, from Abstract Expressionism to 3D Game Art to Aboriginal Dot Painting. Where the other nodes in the pack each specialize (cameras, artists, materials), this is the everything drawer. You pick a style, type your subject, and it welds the style phrasing onto your prompt at a heavy weight. If you just want to throw a look at a scene and see what sticks, this is the one to grab.

That breadth is the appeal and the catch. A 1,500-entry menu is fun to scroll, but a lot of those entries are near-duplicates or ultra-niche, and the ones tagged _Uncategorized (like Aardman_Uncategorized) are exactly what they sound like - grab-bag additions with no promise of doing much. Treat it as a slot machine you can afford to pull, not a curated shortlist.

How it works

MileHighStyler is a maximalist fork of the classic SDXL Prompt Styler - the drop-down style-injection node plenty of people already run (the phrase "prompt styler" shows up constantly in ComfyUI discussion for a reason). Each of the 1,500 entries is a JSON template with a {prompt} slot. MilehighStyler drops your text_positive into the chosen style's template and, per the pack docs, wraps everything except your own words at a strong default weight (~2.0) so the style leads.

It's a pure text tool. It rewrites your prompt strings; your model and sampler do the actual generating.

The inputs and outputs that matter

• milehigh - the ~1,500-entry style dropdown. no style is the off switch.
• text_positive - your subject and scene; it lands inside the template.
• text_negative - optional negative text to carry along.

log_prompt prints the finished prompt to the ComfyUI console, which is genuinely useful here - with 1,500 opaque style names, the log is the only way to see what a given pick actually injects. Outputs are text_positive and text_negative strings, straight into a CLIP Text Encode.

Installing it

ComfyUI Manager: search ComfyUI_MileHighStyler, install, restart. Or terminal:

cd ComfyUI/custom_nodes && git clone https://github.com/TripleHeadedMonkey/ComfyUI_MileHighStyler

then restart. No models, no requirements.txt, no dependency drama - Python plus a folder of JSON. Nodes appear under ali1234/stylers. If you want to add your own styles, drop a JSON file into the pack's data/ folder (the README walks through the format) and restart.

Where people get burned

• Model era, first and loudest. The (style:2.0) weighting this bakes in is CLIP-encoder syntax. On the 2026 LLM-encoded models - Flux 2, Z-Image, Qwen, Anima, Krea - weights are silently discarded and the punctuation is read literally, so the node's whole mechanism is inert. Run it on the SDXL family (SDXL, Illustrious, Pony, NoobAI, WAI) or SD1.5. On newer models, describe the style in plain words instead.
• Style bulldozes subject. With that heavy default weight, a strong style can flatten your composition. Front-load concrete subject detail and don't stack a second styler until you've seen what one does.
• Menu paralysis. 1,500 options is more curse than blessing. Pick a handful you trust and stop scrolling; most entries won't move your particular checkpoint anyway.
• CFG 1 kills the negative half. Any negative an entry carries only does something above CFG 1 - on Turbo/Lightning checkpoints it isn't computed.