Everyone obsesses over face consistency — correctly — but the second-most-common tell that an account is "AI slop" is sartorial: a character with infinite clothes and no taste. Real people have closets — finite, repeating, coherent. Giving your character one is cheap and changes how the whole feed reads.
Why wardrobe drift breaks immersion
Followers don't consciously audit outfits, but they absorb patterns. A human posts the same leather jacket forty times a year; her style has a center of gravity. A character wearing a brand-new, perfectly-styled outfit in all 90 posts reads as a catalog — beautiful, and obviously nobody. The realism layer isn't just sensor grain; it's repetition.
There's also a practical failure: without wardrobe rules, every prompt re-invents her style from scratch, and the feed's aesthetic wanders the same way un-anchored faces do.
Build the closet: a one-page document
Like the backstory voice card, this is half a page you'll use forever:
- 3–4 signature pieces that recur explicitly: "the oversized charcoal blazer," "the silver chain she always wears," "beat-up white sneakers." Signature pieces are continuity anchors and content hooks (followers genuinely comment when the blazer returns).
- A palette: 4–5 colors plus one accent. Every outfit prompt pulls from it; the grid coheres automatically.
- Fit rules: "oversized top, slim bottom," "never bodycon," "always layered." Three rules generate infinite on-brand outfits.
- No-list: styles she'd never wear. As much identity as the yes-list.
Then prompt from the closet: "her usual oversized charcoal blazer over a white tee, gold hoops" — written once, reused everywhere. (The garment-description craft is the one place verbose prompts help, since clothing is scene, not identity.)
The reference-library angle
Wardrobe has a pipeline dimension too: the golden set should include the character in her typical clothing register, including fitted pieces — that's what teaches generation her actual silhouette rather than a generic body under interchangeable outfits. If her reference library is all evening gowns and your daily content is gym fits, you're asking the model to extrapolate body shape every time, which is where drift sneaks in.
Repetition mechanics for the feed
- Re-wear deliberately. Prompt the same signature piece across different scenes in the same month. This is the single cheapest "she's a real person" signal that AI accounts skip.
- Seasonal rotation. Closets change with weather; introduce 2–3 "new" pieces per season and retire others. It creates natural content moments ("first wear of the season") and keeps the palette from fossilizing.
- One wildcard lane. A defined exception — "she goes full glam for events" — gives you variety inside a rule, which reads as range rather than randomness.
Exact-garment expectations
Text prompts reproduce types of garments reliably, not specific SKUs — "the charcoal blazer" will be consistently a charcoal oversized blazer, with cut details varying slightly between renders. Camera distance is your friend (mid and full-body shots make minor garment variance invisible); detail crops of the "same" item are where eagle-eyed followers could spot differences, so build those from your best single render rather than regenerating.
A character with a closet, a palette, and a no-list feels authored. It takes thirty minutes to write — right after the face takes ten.