🗒️LC Scene Builder

Every photographer-turned-prompt-writer knows the grind: describing an environment well takes a paragraph, and doing it from memory means it comes out thin. LC Scene Builder is the environment half of the LC123 prompt-builder family, and it exists so "where does this happen" is a dropdown instead of a wall of text you have to get right.

Pick a scene preset - 25 of them, from "Luxury penthouse" and "Modern kitchen" to "Rain-slick street" and "Alley with graffiti" - and the node fills in a detailed environment description for you. Then add time of day, weather, and any props or atmosphere the preset doesn't cover. What comes out the prompt socket is a ready-to-use STRING that feeds the LC Prompt Assembler (or any text encoder).

How it works

The presets come from editable JSON files in assets/prompt_builder/, so the list isn't hardcoded - you can tweak it. The mechanism has one clever bit worth knowing: it strips baked-in time and weather phrases from a preset when you've supplied your own in the widgets. Pick "Golden hour" in the dropdown and a preset that says "sunset-lit terrace" won't fight you with doubled phrasing. It then appends its own time/weather detail lines, joins everything with commas, and tacks on machine-readable scene-object markers (|||LC_OBJ:...|||) that only the assembler cares about - the caption text itself stays clean.

The inputs

• scene_preset - the environment preset (default "Luxury penthouse"). This fills the scene description.
• description - freeform text to override or extend the preset; empty = preset only.
• time_of_day - none/Dawn/Morning/Midday/Afternoon/Golden hour/Dusk/Evening/Night. "none" leaves time out.
• weather - none/Clear/Partly cloudy/Overcast/Foggy/Rainy/Snowy/Windy.
• environment_details - props, architecture, atmosphere not covered by the preset.

Output is one prompt STRING.

Where it fits

The README's stack is: Subjects + Scene + Camera + Lighting + Style + Palette → LC Prompt Assembler → prompt → CLIP. So Scene Builder is one module in a modular build - pair it with LC Subject for the character, LCStyleSelector for the look, and the assembler stitches it together. You can also just wire it straight into a CLIP Text Encode if you want a fast, detailed environment prompt without the rest of the machinery.

One honest note: on tag-based models (Illustrious, NoobAI, Pony lineage) full descriptive sentences can dilute adherence versus proper booru tags; the preset's prose works best on LLM-encoded models that read natural language (Flux/Klein/Anima families) - which is the audience this pack is clearly aimed at.

Install

Part of LC123 - ComfyUI Manager (search "LC123") or:

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

Restart. No extra pip dependencies; MIT licensed. Cloud-hosted ComfyUI needs LC123 requested or run locally.