Breast_State Styler

Breast_StateStyler is a focused little node: a dropdown of ~91 anatomy and coverage tags - cleavage, sideboob, underboob, cleavage cutout, natural shape, and so on - that it injects into your prompt. It's one of the more-clicked nodes in this pack, which tells you what a lot of people are actually using MileHighStyler for. No judgment here, just accuracy: it's a menu that saves you from typing (and misspelling) NSFW booru tags by hand.

The single most useful thing to understand about it is that these are Danbooru tags. They only mean something to a model that was trained on Danbooru-tagged data - the anime/NSFW SDXL finetunes like Pony, Illustrious, NoobAI, and WAI, plus the older SD1.5 anime lineage. Feed the same tag to a photoreal checkpoint or SDXL base that never saw booru captions and you'll get little or nothing. The dropdown can't know what your model learned; it just pastes the tag.

How it works

MileHighStyler is a fork of the classic SDXL Prompt Styler. Each menu entry is a JSON template with a {prompt} slot; Breast_StateStyler drops your text_positive into the chosen tag's template and wraps the style text at a strong default weight (~2.0, per the pack docs). Output is text - it flows into your CLIP Text Encode. Nothing here touches the model or the image; it rewrites your prompt.

The inputs and outputs that matter

• breast_state - the ~91-entry tag dropdown. none is the off switch.
• text_positive - your subject and scene; it lands in the template.
• text_negative - optional negative text.

log_prompt prints the finished prompt to the console - worth leaving on so you can see exactly which tag got added and at what weight. Outputs are text_positive and text_negative strings, into a 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 model files, no requirements.txt, no heavy dependencies - Python and JSON. Nodes appear under ali1234/stylers.

Where people get burned

• Wrong checkpoint. These are booru tags. On a model that wasn't trained on booru data, they're inert. Match the node to an anime/NSFW SDXL finetune (Pony, Illustrious, NoobAI, WAI) and it works; on photoreal or base SDXL, it mostly won't.
• Model era. The 2.0 weighting is CLIP-encoder syntax and gets discarded on the 2026 LLM-encoded models (Flux 2, Z-Image, Qwen). Anima, being booru-trained but LLM-encoded, may know the tag but ignore the weight - so on those, write the tag plainly. This node is happiest on the SDXL/SD1.5 CLIP family.
• Over-weighting anatomy. Because the tag is wrapped heavily, it can dominate the pose and framing. If the composition goes sideways, that heavy weight is usually why - front-load your actual subject.
• Contradictory tags. Stacking coverage tags that describe incompatible states just confuses the model. Pick one.