AI Virtual Try-On
Put a garment on a person: upload the photo and the piece, and see it worn with the right fit, the right drape and the same face and pose.
- Two photos
- Same face and pose
- Indian wear supported
- No watermark
Free credits to start. No watermark.
Virtual try-on answers a question that a product photo cannot: how would this look on a body like mine? Two images are enough — a person standing straight and the garment laid flat or on a hanger — and the model fits one to the other, keeping the face, pose and lighting of the original photo.
It is the difference between a shopper guessing and a shopper deciding. For a brand, it is the difference between one garment photo and a full size or colour matrix without booking a shoot for each variant.
What two photos you need
| Photo | What good looks like | What breaks it |
|---|---|---|
| Person | Standing straight, plain background, full or half body, even light | Arms crossed over the chest, seated, heavy shadow |
| Garment | Flat lay or hanger, whole piece visible, plain background | A cropped sleeve, a busy print behind it, a model already wearing it |
A garment photographed on a person is harder to transfer than a flat lay, because the model has to separate the garment from that body first.
How to try on clothes in three steps
- 1Upload the personJPG, PNG or WEBP up to 4 MB, standing straight, plain background. The photo becomes the anchor for face, pose and light, so a clean one pays off immediately.
- 2Upload the garmentA flat lay is best. The colour, print and cut are copied from this image, so make sure the whole piece is in frame and the lighting is neutral.
- 3Choose type and fitTop, dress, bottom, saree or jacket, then fitted, regular or relaxed. The fit setting decides whether the garment follows the body or drapes away from it.
Sarees, kurtas and the drape problem
Indian wear is where most try-on tools fail, because a saree is not a shape, it is a construction: pleats at the waist, a pallu over one shoulder, a fitted blouse. The prompt here states that construction explicitly, which is why sarees preview well on a straight standing pose and badly on a moving one.
Kurtas and sherwanis are more forgiving, since the silhouette is closer to a long shirt. Where the details still fail is embroidered borders and fine prints, which can be reduced to generic texture. If the exact embroidery matters, check it at 100 percent and regenerate rather than accept a softened version.
Where brands use try-on
- Ecommerce: show one garment on several body types without a shoot per variant.
- Colour variants: same pose, same person, five colours, one afternoon.
- Pre-order campaigns: show a sample on a model before production.
- Styling previews: test whether a piece works with an existing wardrobe.
- Wedding shopping: see a lehenga or sherwani on the actual person before the fitting.
Honest limits
Fit is an approximation, not a measurement. The render shows how a garment might sit, not whether the size will fit, and it cannot tell you about fabric weight or stretch. For anything expensive, treat the image as a styling preview and confirm size with the brand.
The other limit is the reference garment. If the flat lay is blurry, poorly lit or partly out of frame, the transferred garment inherits those problems. A clean reference is worth more than a stronger model.
Credits, formats and privacy
Each try-on costs credits, priced per model and shown before you run, and new accounts start with free credits. Output is 4:5 at up to 1344 pixels, no watermark. Both uploads are capped at 4 MB.
Images are sent only for the duration of the request and the result is stored so you can download it, with no public gallery and no email capture. Do not upload photos of other people without their consent.
Lighting and background: the two things that sell the try-on
A try-on is believable when the garment is lit like the person, not like the flat lay it came from. If the person was photographed in window light from the left, the fabric has to carry that light on its left side, and the shadows in the folds have to match the shadows on the face. That consistency is what the model is being asked to solve, and it is why a person photo with one clear light direction works so much better than a photo with mixed indoor lighting.
The background matters for a different reason: a busy background competes with the garment and hides the edges. A plain wall, a studio backdrop or the Remove Background tool's white output all give the render somewhere clean to land, and the result is an image you can put straight on a product page.
Who this replaces, and who it does not
It does not replace a fit model, because a fit model tells you how a garment behaves when a person moves, sits and raises an arm. It replaces the awkward middle: the pre-launch image, the colour variant, the styling test, the “will this work with what I own” question that otherwise costs a return.
Consent and the limits of a borrowed photo
A try-on uses someone's body and face, which means it needs their agreement, in the same way a photograph does. For a brand this is a model release; for personal use it is a conversation. Generating a try-on of someone who has not agreed is not a technical shortcut, it is a consent problem.
Testing a size range without a shoot
Run the same garment on three different body photos rather than generating dozens of variants, and you will see where the fit breaks: shoulders, waist or length. That is the information a size chart cannot give you before a shoot, and it costs three generations.
One last check before you publish
Look at the hem, the collar and the sleeve length in that order. A garment that ends in the wrong place, or a collar that floats above the neck, is what makes a try-on look generated rather than photographed, and those three details are visible even at thumbnail size.