When to use this tool
- Shop listings on a human model without a shoot
- Lookbooks and outfit previews
- Marketplace listings for a single garment
- Checking a purchase against a product photo
AI fashion tool
Virtual Try-On
Upload two photos — the person, then the garment on its own — and get the same person wearing it. The model reproduces the garment’s colour, fabric, pattern, seams and length, and is asked to leave the face, hairstyle, pose, body proportions and background exactly as they were.
2 images · JPG, PNG or WebP up to 10 MiB each · about a minute · 1 credit per task
How it works
- Upload the person — The first image is the subject: a JPG, PNG or WebP of the person, ideally full body and facing the camera. The output keeps its shape.
- Upload the garment — The second image is the garment on its own — a product or flat-lay photo works best. The order matters: person first, garment second.
- Say what it covers — Choose a whole outfit, a top, trousers or a skirt, or a dress, then run the task. It takes about a minute and uses 1 credit.
Frequently asked questions
What do the two images have to be?
The first is the person, ideally standing and facing the camera with even light; the second is the garment on its own, like the product shot from a shop. The order is fixed — the first image is the body, the second is what it wears.
Does it keep my face and body?
The prompt asks the model to keep the face, hairstyle, skin tone, body shape, pose and background unchanged, and to change only the clothing. Because it re-renders the whole frame, small differences in hair, hands or background detail are possible, so review the result at 100% before you use it.
Can it show a whole new outfit?
One garment reference is read per task, so pick what that garment covers: a top keeps the trousers or skirt, a dress replaces the whole outfit, and the outfit setting treats the reference as a complete top-and-bottom set. Two separate pieces need two tasks.
Is there a size limit on the images I upload?
Each file is capped at 10 MiB and must be a JPG, PNG or WebP. There is no pixel ceiling here — the references are scaled down inside the service before the frame is rendered.