Breast_State Styler (Advanced)

This is the SDXL dual-CLIP version of Breast_StateStyler: the same ~91 anatomy and coverage tags - cleavage, sideboob, underboob, cleavage cutout, natural shape - but with the positive prompt split across SDXL's two text encoders and a control for where the negative lands. It carries one of the higher click-through rates in the pack, which is a fair signal of what a lot of MileHighStyler users are here to do. Described plainly, because that's what it is: a dropdown that saves you typing NSFW booru tags.

The one fact that decides whether it works at all: these are Danbooru tags, so they only mean something on booru-trained models - the anime/NSFW SDXL finetunes (Pony, Illustrious, NoobAI, WAI) and the SD1.5 anime lineage. On a photoreal checkpoint or SDXL base, most do nothing.

How the "Advanced" plumbing works

SDXL has two text encoders, CLIP-G and CLIP-L. This node feeds them separately - text_positive_g to G, text_positive_l to L - and applies the tag to both, wrapped at the pack's strong default weight (~2.0). MileHighStyler is a fork of the classic SDXL Prompt Styler. negative_prompt_to (Both / G only / L only) routes the entry's built-in negative; Both is the sane default.

The inputs and outputs that matter

• breast_state - the ~91-entry tag dropdown; none is off.
• text_positive_g / text_positive_l - your subject per encoder. Same text in both if unsure.
• text_negative - your negative, with the style's negative routed by negative_prompt_to.

log_prompt prints the assembled prompt so you can see the exact tag and weight. Six outputs: text_positive_g, text_positive_l, merged text_positive, and text_negative_g / text_negative_l / merged text_negative. Wire _g / _l into a CLIPTextEncodeSDXL, or take the merged text_positive / text_negative into 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.

Where people get burned

• Wrong checkpoint. Booru tags need a booru-trained model. On photoreal or base SDXL they're inert. Match to Pony/Illustrious/NoobAI/WAI.
• Advanced when basic would do. The G/L split only earns its keep with a CLIPTextEncodeSDXL and a real reason to differ per encoder. Otherwise use the merged outputs, or the plain node.
• Model era. Dual-CLIP is SDXL-only, and this node is meaningless on the 2026 LLM-encoded models (Flux 2, Z-Image, Qwen). Even a booru-trained LLM-encoded model ignores the weight - write the tag plainly there.
• Over-weighted anatomy hijacks the pose. If framing goes odd, the heavy tag weight is usually the cause. Front-load your actual subject.