Outfit Generator

Describing a coherent outfit in a prompt is more finicky than it sounds - the model hears "red silk blouse, leather skirt, heels" and quietly decides which parts to actually render. Outfit Generator takes that guesswork out of your hands: it assembles an outfit piece by piece from a curated garment database and emits a properly ordered description, seeded so the same seed gives the same outfit every time. It's the piece of the FVMtools color/fashion pipeline that feeds Prompt Color Replace - it writes the garments with #color# placeholders, and Prompt Color Replace fills in the actual palette colors afterward.

How it works

The outfit "database" is not a database - it's a folder of plain-text files. Each outfit set lives in outfit_lists/<set>/ with up to ten .txt files (top, bottom, footwear, headwear, outerwear, accessories, bag, fabrics, plus optional prints and texts). Each line in a garment file follows a format like:

blouse | 0.65 | 0.3-0.7 | silk,chiffon,cotton,satin,lace

garment name, selection probability, a formality range, and candidate fabrics. The node rolls for each garment, filters by your formality slider (a blouse eligible for 0.3–0.7 only appears when your formality sits in that range), picks a fabric, and optionally adds a print or text based on print_probability.

The bundled set list is big - the dropdown alone exposes a few hundred entries, spanning business, casual, evening, aerobic, beach, streetwear, sheer/layered, lingerie, and more, organized into female//male/ and activity subfolders - and it's dynamically scanned from the folder. The killer feature is the Edit List button: it opens the source files in the browser, you edit them, and changes take effect on the next Queue run. No restart, no reinstall. You can drop a whole new custom set into outfit_lists/ and it appears in the dropdown instantly.

The inputs that matter

• outfit_set - which collection to draw from. general_female is the default; the dropdown lists everything in outfit_lists/.
• seed - same seed + same settings = same outfit. Change it for variety.
• formality and coverage - the two sliders that shape the result. Formality filters garments by their formality range (0 = casual, 1.0 = tuxedo). Coverage decides how many optional slots (headwear, outerwear, accessories, bag) get activated - top/bottom/footwear are always on when enabled.
• print_probability - chance each garment gets a print/pattern/text (0.3 default).
• text_mode - how text on clothing appears: quoted writes exact text like "REBEL" text in gothic font (best for text-aware models like Z-Image Turbo and Flux2), descriptive writes a generic "bold text graphic" (safe for SD/SDXL which can't render text), off skips text.

Three outputs: outfit_prompt (the tag-filled description for Prompt Color Replace), outfit_details (a structured breakdown of what was picked), and outfit_info (full generation metadata). There's also an override_string optional input - one line per slot, top: silk blouse | #primary# bypasses generation entirely for that slot, which is the precise way to fix one garment without regenerating.

Installing

Same pack, zero extra dependencies:

cd ComfyUI/custom_nodes
git clone https://github.com/ping1979ping/comfyui-FVMtools

Restart ComfyUI. No models needed - this corner of the pack is pure text generation.

Where people get confused

Two things. First, formality must match your chosen set - a yoga set whose garments live in the 0.0–0.2 range will produce almost nothing at formality 0.5. The tooltip says it plainly: match the slider to the set. Second, remember the tags are placeholders - feeding the raw outfit_prompt straight into CLIPTextEncode gives you #primary# as literal text in your prompt, which the model will render as some kind of weird token. Always run it through Prompt Color Replace (or set text_mode appropriately and wire in your own colors) before it hits the encoder.