Milehigh Styler (Advanced)

The pack's big catch-all style menu - ~1,500 entries, everything from Abstract Expressionism to 3D Game Art - in its SDXL dual-CLIP form. If MilehighStyler is the everything-drawer, this is that same drawer wired for SDXL's two text encoders, with the positive prompt split into a G stream and an L stream and a knob for where the negative goes.

Be honest with yourself about whether you need the "Advanced" bit. A 1,500-style menu is already a lot to manage; adding per-encoder text routing on top is a power-user flourish most people will never touch. If you just want to fling a style at a scene, the plain MilehighStyler does it with far less wiring.

How the "Advanced" plumbing works

SDXL runs two text encoders, CLIP-G and CLIP-L. This node addresses them independently: text_positive_g to G, text_positive_l to L. Whichever of the ~1,500 styles you pick is applied to both and wrapped at the pack's strong default weight (~2.0) - MileHighStyler being a fork of the classic SDXL Prompt Styler. negative_prompt_to (Both / G only / L only) decides which encoder receives the style's built-in negative; default Both.

It's prompt assembly, not image work. Text out, into your CLIP Text Encode.

The inputs and outputs that matter

• milehigh - the ~1,500-entry style dropdown; no style is off.
• text_positive_g / text_positive_l - your subject per encoder; same text in both if you're unsure.
• text_negative - your negative, with the style's negative routed by negative_prompt_to.

log_prompt prints the assembled prompt - essential here, since 1,500 opaque names give you no other way to know what a pick injects. Six outputs: text_positive_g, text_positive_l, merged text_positive, and text_negative_g / text_negative_l / merged text_negative. Feed _g / _l into a CLIPTextEncodeSDXL, or just use the merged text_positive / text_negative in a normal CLIP Text Encode.

Installing it

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

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

then restart. No models, no requirements.txt, no dependencies - Python and JSON. Nodes appear under ali1234/stylers. You can add your own styles by dropping JSON into the pack's data/ folder (README has the format).

Where people get burned

• Advanced when basic would do. The G/L split only matters if you're driving a CLIPTextEncodeSDXL with deliberately different per-encoder text. Otherwise grab plain MilehighStyler.
• Model era. Dual-CLIP is SDXL-only; this node is meaningless on the 2026 LLM-encoded models (Flux 2, Z-Image, Qwen), which lack a G/L pair and discard weighting. It's SDXL/SD1.5-family by design.
• Style bulldozes subject. With ~2.0 weighting, a strong style flattens composition. Front-load subject detail; don't stack.
• Menu paralysis. Many of the 1,500 are near-duplicates or won't move your checkpoint. Find a handful that work and stop scrolling.