Azazeal Styles Styler (Advanced)

Same 425-entry punk/dystopian/album-art menu as the regular AzazealStylesStyler - anarcho-punk, grunge, zine covers, vinyl art - but this is the "Advanced" variant, which means one thing: it's built for SDXL's two text encoders. If you don't know what that means, you probably want the plain node. If you do, this gives you finer control over how the style lands.

The one thing that makes it "Advanced"

SDXL doesn't have one text encoder, it has two - a big OpenCLIP-G and a smaller CLIP-L - and it runs both, combining their output. Most nodes hide that and feed both the same text. This node splits it open: instead of one text_positive, you get text_positive_g and text_positive_l, and you decide what each encoder reads. The usual play is broad/conceptual text to G and concrete detail to L, but honestly, feeding both the same thing is fine too - the split is optionality, not homework.

negative_prompt_to is the other advanced knob: a three-way choice of Both / G only / L only that picks which encoder your negative prompt hits. Leave it on Both unless you're chasing a specific artifact.

None of this matters off SDXL. On SD 1.5 there's only one encoder; on the 2026 LLM-encoded models (Flux 2, Z-Image, Qwen) the whole CLIP-G/CLIP-L world doesn't exist - the split is dead weight and you should use the basic node, or skip the pack entirely.

How it works

It's a fork of the classic SDXL Prompt Styler. Each Azazeal entry is a JSON template with a {prompt} slot; your text drops in, the style text around it gets wrapped at a strong weight (~2.0), and the node emits rewritten prompt strings. It's a text rewriter - nothing loads a model.

Inputs and outputs that matter

• Azazeal Styles - the 425-style dropdown; None = off.
• text_positive_g / text_positive_l - your subject, one copy per SDXL encoder.
• negative_prompt_to - Both (default), G only, or L only.

log_prompt prints the built prompt to the console. The node hands back six strings: text_positive_g, text_positive_l, and a combined text_positive, plus the same three on the negative side. Wire _g and _l into the matching halves of an SDXL dual CLIP Text Encode - or just grab the combined text_positive / text_negative and treat it like the basic node.

Installing it

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

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

No models, no dependencies - pure Python and JSON. Nodes land under ali1234/stylers.

Where people get burned

• Using it off SDXL. The G/L split is the entire point of the Advanced node, and it only means something on SDXL. On anything else, you're wiring extra sockets for no reason.
• Overthinking G vs L. People agonize over which encoder gets what. Start with the same text in both, ship, and only split them if you have a concrete reason. It's a nudge, not a science.
• Strong weight, strong menu. 425 Azazeal styles at 2.0 will happily bulldoze your subject. If it's too much, that's the default weight - stack a lighter styler after it or pick a subtler entry.