Clothes Changer Selector

Clothes Changer Selector is the wardrobe node in the Multi-Angle Camera Selector pack: a dropdown of 58 outfits that turns into Change clothes to cyberpunk streetwear as one ready-to-wire string. It's the twin of the pack's Background Changer Selector - same custom-field override, same chainable output - but aimed at the character rather than the scene. If you've ever stared at a workflow and wanted to re-dress a subject without rewriting the whole prompt by hand, this is the node for that.

How it works

Pick an outfit, get a prompt clause. The node wraps your choice in the prefix Change clothes to and returns a single STRING named prompt. No inference, no model files, no network - it's a few lines of string building on top of Python's standard library, and that's the whole thing.

The list is genuinely curated rather than padded. It spans formal (three-piece suit, ball gown), casual (white t-shirt and jeans, sundress), workwear (chef whites, firefighter turnout gear), athletic (ski jumpsuit, wetsuit), cultural (kimono, hanbok, lederhosen), historical eras (Victorian gown, 1980s power suit, 1990s grunge), subculture (gothic lolita, steampunk attire), fantasy (knight's armor, space suit), and loungewear (pajamas, silk robe). Fifty-eight distinct looks is a real lookbook, not filler.

The important detail, same as its sibling: the custom_clothes field overrides the dropdown whenever it's non-empty. Type hand-knitted sweater and corduroy and that's what goes into the prompt, dropdown be damned, seed be damned. The presets are a starting point; the free-text field is where you actually live once you outgrow them.

Inputs and outputs that matter

• clothes - the 58-outfit dropdown plus random. Default business suit.
• custom_clothes - free text, multiline, wins over the dropdown when non-empty.
• seed - only does anything with random selected and the custom field empty.
• previous_prompt (optional) - daisy-chain from another selector.
• Output: one STRING named prompt, into CLIPTextEncode or the next selector's previous_prompt.

Chain a full character beat in one sentence: Change clothes to noir trenchcoat look, Change facial expression to smirk, Change pose to leaning against wall. That instruction-style phrasing is aimed at Flux 2 Klein 9B, the editing model in the pack's name, which wants short "change X to Y" commands - though the output works as plain prompt text on anything.

Install

ComfyUI Manager, search "Multi-Angle Camera Selector", install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/saluzafa/comfyui-klein9b-multi-angle

Same story as the rest of the pack: no dependencies beyond the standard library, no models to download, MIT licensed. It sits under prompt/character in the node menu.

Gotchas

• The custom-overrides-dropdown behavior cuts both ways. If the dropdown seems frozen on an outfit you don't remember choosing, check custom_clothes - there's likely an old value in it.
• It writes the words, it doesn't stitch the garment. On a Klein editing workflow the "change clothes to X" instruction is exactly what the model expects; on a plain checkpoint the phrase just conditions the render, so outfit adherence varies. When it doesn't stick, that's model behavior, not a node bug.
• random + randomized seed = a free rotating wardrobe for character reference batches. Same person, different outfits, no hand-editing - the bracket-trick idea from classic prompt tooling, minus the syntax.