Virtual try-on: how AI puts clothes on your own photo

How it works underneath

Older virtual fitting rooms warped a flat garment image onto a detected body pose. The result looked pasted on, because it was. The 2026 approach is different: a generative image model receives your photo and the garment photos together and re-renders the whole frame, so the clothing is generated in the scene's own lighting, with real folds and shadow.

That is why the input photo matters so much. The model is not overlaying — it is reasoning about your body, your pose and the light you were standing in, and drawing all of it again.

How accurate it is, honestly

• Silhouette and drape: reliable. A blazer will hang like a blazer.
• Colour: reliable when the garment photo is well lit.
• Fit on your body shape: good, if your reference photo is full-length and front-facing.
• Fine print, slogans and logos: unreliable. Text on garments is the single weakest area.
• Complex textures — sequins, sheer layers, intricate lace: variable.
• It shows you how something looks, not how it feels or how it sits in the shoulders. It reduces returns; it does not eliminate them.

Taking a reference photo that works

Stand facing the camera, full body in frame from head to shoes, arms slightly away from your sides, against a plain wall in even daylight. Wear fitted clothing — a baggy hoodie hides the body line the model needs to infer. One good photo is worth more than ten mediocre ones.

In AI Wardrobe you can save several photos and mark one as active; the active photo is the one every try-on uses, so update it when your hair or body changes noticeably.

Quick mode versus quality mode

Quick mode returns a result in a few seconds at lower resolution — right for scanning several options. Quality mode runs a stronger image model at 1K resolution and preserves garment detail far better; use it for the look you are actually deciding on.

Trying on things you don't own yet

Try On Studio lets you upload garment photos from anywhere — a shop page, a screenshot, a friend's photo — and see them on yourself before buying. The browser extension does the same in one click from a product page, and saves the item to your wishlist.

Your photos are stored in private buckets and served through short-lived signed URLs. They are not used to train models and are not shared.