Describe This Outfit

A useful outfit description is structured, not poetic. It names each piece by its real garment type — a "wrap dress", not "a dress"; "straight-leg raw-hem jeans", not "pants" — because garment vocabulary is what buyers search and what stylists communicate in. Then it pins down color with precision (sage rather than green, oxblood rather than red), names any print or pattern, and reads the fabric from its visual texture: ribbed knit, brushed wool, washed satin, distressed denim.

Fit and silhouette come next, because two identical garments in different cuts are different outfits: oversized versus tailored, cropped versus longline, high-waisted versus mid-rise, A-line versus bodycon. Then the description zooms out to the aesthetic the pieces add up to — minimalist, streetwear, quiet luxury, cottagecore, business casual — and the occasions the outfit actually suits, which is the sentence most people find hardest to write themselves.

The AI here follows exactly that structure, in order, every time. It describes only the clothing — never the person wearing it — flags materials it cannot confirm from pixels with an honest "appears to be", and tells you when part of the outfit is cropped out of frame instead of inventing it. That last part matters more than it sounds: a description you can trust not to hallucinate a fabric is one you can paste into a listing without re-checking every claim.

Dresses are the hardest single garment to describe well, and the one where description quality most visibly moves a sale. A dress listing that says "blue midi dress" competes on price alone. One that says "cornflower-blue satin slip midi with a cowl neckline, adjustable spaghetti straps and a side slit, bias-cut for a fluid drape" ranks for the searches real buyers type, answers the questions that otherwise become messages ("what's the neckline like?", "is it stretchy?"), and signals a seller who knows what they are selling.

The vocabulary is the bottleneck. Necklines alone span sweetheart, cowl, square, halter, boat, off-the-shoulder; sleeves run bishop, puff, raglan, dolman, flutter; lengths and cuts have their own taxonomy from mini to maxi, sheath to fit-and-flare. Most sellers know the dress they are holding is lovely and do not have those words on tap. Upload the photo, choose the resale purpose, and the description comes back with the neckline, sleeve, length, cut, fabric and closure named — ready to paste into a Poshmark, Depop, Vinted or eBay listing and then edit with the facts only you know (brand, size, condition, measurements).

Two honest limits. The AI describes what is visible: it cannot read a care label from a distance shot, verify fiber content, or know the label size — add those yourself, and photograph the tag if you want it described. And marketplace buyers forgive a plain description sooner than a wrong one, which is why the tool is instructed to hedge ("appears to be silk") rather than assert what the pixels cannot prove.

Image alt text is the description a screen reader speaks to a blind or low-vision user in place of the photo. For fashion content — shops, blogs, social posts — good alt text is genuinely hard: it has to convey the outfit's essentials in a sentence or two, without the filler ("image of a photo showing...") that wastes a listener's time, and without editorializing about the person in the picture.

The alt-text purpose setting writes for exactly that constraint: garments, colors and the overall look, compressed, front-loaded and neutral. Pair it with the brief detail level for a single-sentence alt attribute, or detailed when the image is the content itself — a product page hero, a lookbook slide — and merits a fuller treatment.

Beyond compliance, alt text is one of the few places where accessibility and SEO pull in the same direction: descriptive, specific alt text helps image search understand fashion photos the same way it helps screen-reader users. Writing it by hand for a 200-product catalog is the kind of task nobody finishes; generating a draft per image and editing the exceptions is the version that ships.

The styling-notes mode reads an outfit like a stylist's file card: what the pieces are, how they are layered and proportioned, what the styling choices are doing (the tucked-in knit, the rolled cuff, the belt breaking the silhouette), and what the look would pair with. It is the fastest way to build a written reference from street-style photos, fitting-room mirror shots or your own wardrobe — searchable notes instead of a camera roll you scroll forever.

The research mode leans analytical: aesthetic classification, the era or movement a look draws from, construction details worth noting. Students documenting looks, vintage sellers dating a garment's style cues, designers cataloging references and writers describing runway or archive images all need the same thing — accurate, structured language about clothes — and the additional-questions box turns each analysis into a targeted one: ask "what decade does this dress evoke?" or "what would substitute for the jacket?" and the answer comes back alongside the description.

One boundary worth stating: the tool describes clothing, not people. It will not comment on bodies, guess measurements, or evaluate how someone looks — it treats every photo as an outfit to be documented, which is what makes it usable for accessibility work and professional cataloging alike.