How Try-On Works & Limitations
What the image generation in TryOn does, what it is for, and where its results fall short.
A plain description of what happens when you tap Generate, what the result is good for, and where it goes wrong.
What it does
TryOn takes two kinds of image — a photo of a person and one or more garments — and produces a new image of that person wearing those garments. The result is generated: an AI image model draws it. It is not a cut-and-paste of the garment onto your photo, and it is not a 3D simulation of fabric on a measured body.
Generation runs on our servers using a third-party AI image-processing provider. It receives the images and nothing that identifies you, and it does not use them to train models. The Privacy Policy covers this in full.
What it is for
Helping you picture a look — a colour against your skin tone, a silhouette, how pieces go together — before you buy, pack or get dressed. It is a visual aid for personal use.
What it is not
- Not a sizing tool. The model does not know your measurements or the garment's. It draws clothes as fitting, whether or not the real item would.
- Not a product photo. Do not rely on a result for the exact colour, fabric, length or pattern of a real garment.
- Not a record of reality. Every result is a synthetic image.
Known limitations
- Fit and length are the least reliable part. Sleeves, hems and waistlines can come out longer, shorter, looser or tighter than the real garment.
- Fine detail drifts. Small text, logos, prints, stripes and intricate patterns can blur, change or disappear.
- Hands, hair and faces can change subtly, especially where they overlap the clothing. Fingers are a frequent problem.
- Backgrounds can pick up artefacts around the edges of the body.
- Layering is harder than a single garment. Each added piece makes a wrong result more likely.
- Unusual inputs work worse: side or back views, seated or cropped poses, several people in the frame, heavy shadows, or a garment photographed crumpled or partly hidden.
- Results vary. The same inputs can give a different image each time. Higher quality settings cost more credits and usually — not always — give a better result.
- Uneven performance. Image models are trained on large collections of pictures that do not represent every body, skin tone, age, disability or style of dress equally well. Results may be less accurate for people and clothing that were under-represented in that data. If TryOn handles your photo badly, tell us — that is what the report button is for.
Getting better results
- A clear, well-lit, front-facing photo of one person, standing, with arms away from the body.
- A garment photo that shows the whole item, flat or on a plain background.
- One garment at a time when accuracy matters most.
Safety
Requests are checked automatically against the Content Policy. A refused request returns no image and costs no credits.
Telling us something is wrong
On any result, choose Report a problem and tick what went wrong — fit, colour or pattern, face or body, hands, background, missing garment. Reports tell us where the model is failing. Your photos are never used for training.
Contact us
Questions about this page, or about TryOn? We usually reply within a day.
- Developer
- Aquison Technologies
- Website
- aquisontechnologies.com