Nodes/ComfyUI LC123 Nodes/πŸ—’οΈLC Style Selector
ComfyUI Node

πŸ—’οΈLC Style Selector

The 'how should this look' module of the prompt builder

By lonecatone23Β·Created 2 months agoΒ·Updated 3 days agoΒ· 18
πŸ—’οΈLC Style Selector
    • prompt
    β—„promptβ–Ί
    β—„style_categoryPhotorealisticβ–Ί
    β—„style_presetUltra-realistic commercialβ–Ί
    β—„quality_levelEditorial qualityβ–Ί

    Describing a style well is harder than describing a subject. "Photorealistic" gets you a starting point; "editorial magazine portrait" gets you a direction; "commercial retouched, hero-grade finish" gets you the polish. LCStyleSelector is the LC123 prompt-builder module that turns "how should this look" into a few dropdowns instead of a paragraph you have to write and re-write.

    It's the style half of the same family as LC Subject (who) and LC Scene Builder (where). You pick a broad style family, a specific preset within it, a quality level, and add any extra notes - film stock, art direction, whatever the presets miss. Out the prompt socket comes a ready-to-assemble style string that feeds the LC Prompt Assembler (or any text encoder directly).

    The inputs

    • prompt - your extra style notes (film stock, art direction, etc.). Optional.
    • style_category - the broad family: Photorealistic, Editorial, Commercial/product, Cinematic, Fashion, Portrait, Documentary, Anime/illustration, Painterly, or Custom.
    • style_preset - the detailed wording within that family: Ultra-realistic commercial, Editorial magazine portrait, Cinematic film still, Fashion lookbook, Anime key visual, Painterly oil study, Product packshot, and more.
    • quality_level - the finish read, from Standard β†’ High detail β†’ Editorial quality β†’ Commercial retouched β†’ Raw/unpolished.

    Output is one prompt STRING. The node joins your notes, the category, the preset's detail wording, and the quality phrasing into a comma-separated style block.

    How it works

    Mechanically it's simple text assembly: prompt + style_category + preset_detail + quality_detail, joined and cleaned. The style presets are editable (they load from assets/prompt_builder/), so you can tune the wording of "Ultra-realistic commercial" to match what your model family actually responds to. The mechanism is honest - it's structured prompt engineering, not a hidden model call.

    Where it fits

    The README's stack: Subjects + Scene + Camera + Lighting + Style + Palette β†’ LC Prompt Assembler β†’ prompt β†’ CLIP. So Style Selector is one module feeding the assembler. A quick take from the KB's prompt-engineering doc: on tag-based SDXL models, flowing descriptive style prose can dilute adherence versus concise tags - this reads best on the LLM-encoded, natural-language families (Flux/Klein/Anima), which is what the pack targets. On those, the four-level structure (family β†’ preset β†’ quality β†’ your notes) produces the kind of layered prompt those models reward.

    Install

    It's in the LC123 pack - ComfyUI Manager (search "LC123") or:

    cd ComfyUI/custom_nodes
git clone https://github.com/lonecatone23/ComfyUI_LC123_nodes

    Restart. No extra pip packages; MIT licensed. Cloud-hosted ComfyUI needs LC123 requested or a local install.

    The take: if you find yourself typing the same style boilerplate into every workflow, this makes the look a reusable dropdown - and the quality-level axis is the part most hand-written prompts forget entirely.

    CategoryLC123/prompt

    Inputs (4)

    NameTypeDefaultDescription
    promptSTRINGExtra style notes (film stock, art direction, etc.).
    style_categoryCOMBOPhotorealisticBroad style family (photo, editorial, anime, etc.).
    style_presetCOMBOUltra-realistic commercialDetailed style wording for this family.
    quality_levelCOMBOEditorial qualityHow polished the finish should read (standard β†’ editorial/hero).

    Outputs (1)

    NameTypeDescription
    promptSTRINGβ€”