Where virtual try-on works, and where it is weaker
Published
Virtual try-on is not equally good at everything, and it helps to know where the limits are before you switch it on across your catalog. This post lists what Aynada supports, what photo each kind of product needs, and the situations where results get weaker, with the reasons behind them.
Which products does Aynada support?
Aynada was designed for clothing, and clothing is its default. Every product is sorted into one of these kinds, and the kind decides how the AI is instructed:
- Clothing, dressed onto the person: tops (T-shirts, shirts, blouses, sweaters, hoodies), bottoms (trousers, skirts, shorts, leggings), one-pieces (dresses, jumpsuits, overalls) and outerwear (jackets, coats, blazers, cardigans, vests), plus a general clothing kind when the exact one is not clear.
- Shoes, which replace the shoes the person is wearing, on both feet.
- Accessories, added to the person while their clothing stays as it is: glasses and sunglasses, hats and caps, bags, necklaces, earrings, bracelets, rings, watches, belts, and scarves, shawls and ties, plus a general accessory kind.
Each kind gets its own rule. A top replaces only the upper-body garment and keeps the trousers; a dress replaces top and bottom; a jacket goes over what the person already wears. Earrings go on each visible ear, a belt at the waist over the clothing, glasses on the nose and ears, replacing any glasses already there.
How Aynada decides what a product is
ikas has no product taxonomy that says “this is a shoe”, so Aynada reads what you wrote: the category names, and the product name when the categories say nothing specific. On Shopify it reads the product type and the product name the same way.
The rules lean towards clothing on purpose. A product with no recognisable category is treated as clothing. So is a product whose categories name both a garment and an accessory, such as “Dresses & Belts”. Two garment kinds, or two accessory kinds, fall back to the general kind unless the product name settles it.
The words Aynada recognises are Turkish and English. A category named only in another language, such as the German “Taschen” for bags, is not recognised, and the product is treated as clothing. If your categories are in another language, check a few products before you switch the button on everywhere.
What Aynada does not offer: underwear and lingerie
A try-on of underwear and lingerie on a shopper’s own photo is not something Aynada offers. On ikas, products recognised as intimate wear stay switched off unless you turn the product itself on. If such a product is switched on and the AI service blocks the image, the shopper is told that virtual try-on isn’t available for it. Swimwear is deliberately not on that list.
What photo does each product need?
The try-on keeps the shopper’s original framing, so the photo has to show the part of the body where the product goes. The ikas try-on window shows a tip that matches the product:
| Product | A good photo | A poor photo |
|---|---|---|
| Clothing | Full-length in a mirror, in good light. Your phone can be in front of your face. | Dark, too close, or the phone covering where the garment goes. |
| Glasses, earrings, hats | Your face and head from the front, sharp and well lit. | Dark, face turned away, or covered by hair or a hand. |
| Shoes | Full-length including your feet, in good light. | Feet out of frame or in the dark. |
| Watches, bracelets, rings | Your wrist and hand sharp and well lit. | Dark, or the hand and wrist covered. |
| Necklaces, bags, belts, scarves and other accessories | Full-length or waist-up in a mirror, in good light. | Dark, too close, or the spot where it goes covered. |
While the image is being created, the window adds a few general tips: even, bright light gives the truest colour, the phone should not cover where the garment goes, and arms held slightly away from the body help.
Where results are weaker, and why
None of the following stops a try-on; each is a reason a result can look less convincing.
- The body part is not in the photo. Aynada does not extend the image or invent parts of the body: if trousers are tried on with a waist-up selfie, only what is visible can change. This is why shoes need the feet in the frame. On ikas, when the browser cannot see the face, the feet or the hand that a product needs, it tells the shopper before they continue.
- Small items in a big frame. A ring, earrings or a bracelet take up very little of a full-length mirror photo. That is why the ikas try-on window asks for a close, sharp photo of the hand or the face for these, rather than the full-length shot that suits clothing.
- Poor light or a covered area. As the window’s own tip says, even, bright light gives the truest colour; a dark photo gives the AI less to go on. A phone, a hand or hair over the spot where the product goes leaves the AI less to work with.
- Busy product images. Product photos often show a model in other clothes. The AI is told to use the product images only as a reference for that one item and to ignore the model, other clothing and background. For jackets it is told explicitly that whatever the model wears underneath is not part of the product.
- An unclear category. A product that falls back to the general clothing or accessory kind gets a general instruction instead of a precise one. Clear category names (“Dresses”, “Sneakers”, “Earrings”) give the AI the more specific rule.
- A long description. On ikas, the AI reads the product’s title, type, categories and the start of its description, up to 500 characters. Size charts and care labels further down do not reach it, so the description’s first lines are where the product itself should be described.
The prompt itself names the mistakes it guards against: the old garment showing under a dress, the reference model’s other clothes being copied, the frame being extended and the body being slimmed. Those rules make the failures less likely; they do not make the result a photograph.
It stays a preview
Whatever the product, the result is an AI preview, not a statement about fit or size, and every result says so. You decide which categories and products show the button, so you can start with the kinds that suit your catalog and add others once you have looked at a few results yourself. For how the try-on works from start to finish, read What is AI virtual try-on?; for what happens to the photo, the post on photo privacy.
Aynada is live on ikas, and Aynada for Shopify is on its way to the Shopify App Store (Shopify page).
Questions and answers
Does Aynada work for shoes and accessories?
Yes. Besides clothing it supports shoes, glasses, hats, bags, necklaces, earrings, bracelets, rings, watches, belts and scarves. Each needs a photo that shows where the item goes.
Can shoppers try on underwear?
No. Try-on of intimate wear on a shopper’s own photo is not offered. On ikas such products stay off unless you switch the product itself on.
Why does a waist-up photo not show the new trousers?
Aynada keeps the shopper’s original framing and does not invent parts of the body that are not in the photo. For trousers, skirts and shoes, a full-length photo is needed.
My categories are not in Turkish or English. Does that matter?
It can. Aynada recognises category words in Turkish and English; a product it cannot place is treated as clothing. Check a few accessories before switching the button on for them.