AI virtual try-on, explained

Virtual try-on used to mean a wobbly 3D jacket floating over a webcam feed. It doesn't any more. Today you hand an app two photos — the garment, and you — and it returns a photograph of you wearing that garment. This guide covers how that works, where it is genuinely useful, and the one thing it still cannot do.

Last updated 16 August 2026 · 9 min read

What virtual try-on actually is

Virtual try-on is any technology that shows a garment on a person without that person physically wearing it. The phrase covers two very different things, and conflating them is why people who tried it in 2019 remember it as a gimmick.

The first kind is AR try-on: a 3D model of a product tracked onto a live camera feed. It works well for objects with a fixed shape and a predictable anchor point — sunglasses on a nose bridge, a watch on a wrist, lipstick on lips. It works badly for clothing, because fabric drapes, and draping depends on a body it has no model of.

The second kind is AI virtual try-on — sometimes called image-based try-on or generative try-on. There is no 3D model and no live camera. An image model takes a photo of the garment and a photo of you, and generates a new photograph: you, in that garment, in your original pose and setting. This is the technique that made try-on useful for clothes, and it is what this guide is about.

How AI try-on works, step by step

From the outside it is two uploads and a wait. Underneath, roughly this happens:

1. The garment is isolatedThe model reads the product photo and works out which pixels are the item — separating a coat from the mannequin, the hanger, the studio floor, and the model wearing it.
2. Your body is readYour photo is analysed for pose, proportions, where your shoulders and waist sit, which way you are facing, and how the light falls on you.
3. The garment is re-rendered onto youNot pasted — redrawn. Sleeves are regenerated to follow your arms, the hem is placed against your actual height, folds and shadows are synthesised to match your lighting.
4. Everything else is preservedYour face, hair, hands, background and pose stay as they were. Only the clothing changes, which is what makes the result read as a photo of you rather than a photo of a model.

The whole round trip takes about 10–30 seconds on current models. That speed is the reason try-on stopped being a novelty: it is now faster than walking to a fitting room.

AR try-on vs AI try-on

	AR try-on / AI try-on
What it needs	AR: a 3D asset built per product. AI: one flat photo of the product.
Catalogue coverage	AR: only products the retailer modelled. AI: anything with a photo, including second-hand listings.
Good for	AR: glasses, jewellery, make-up, furniture. AI: clothing, shoes, layered outfits.
Live or static	AR: live camera, moves with you. AI: a still image you keep and share.
Who has to build it	AR: the shop. AI: nobody — it works on shops that have never heard of the app.

That last row is the practical difference. An AR try-on button only exists on the handful of product pages a brand paid to build. An AI try-on app works on a photo you took yourself in a shop you walked into.

How accurate is it?

Honest answer, split by what you are asking:

• Colour and pattern — reliable. A rust-coloured coat comes back rust-coloured. Stripes stay striped, checks stay checked, and prints land in roughly the right scale.
• Cut and length — reliable. Whether a hem lands mid-thigh or at your knee, whether a neckline is flattering on you, whether a boxy jacket makes you look wider: this is exactly what try-on answers well.
• Drape and texture — good. Heavy wool falls differently from silk, and models handle that convincingly. Very fine detail — a specific weave, small hardware, exact button shape — can drift.
• Sizing — not answered at all. No image model knows your chest measurement. It cannot tell you whether to take an M or an L, and any app claiming otherwise from a single selfie is overselling.

Getting a photo that works

Result quality depends more on your input photo than on which app you use. What helps:

• Full body, or at least head to mid-thigh. A cropped headshot cannot show you a coat.
• Face the camera, arms clear of your torso. Crossed arms and hands in pockets hide the garment lines the model needs to place.
• Even light, plain background. Harsh shadow across your body confuses the lighting match.
• Fitted clothes in the photo. A baggy hoodie leaves the model guessing at your shoulders. A plain t-shirt is ideal.
• Keep two or three photos. One full-body for coats and dresses, one closer crop for tops. Swapping the source photo is usually the fastest fix for a bad result.

The garment photo matters too: a flat product shot or a shot on a model both work, but a crumpled item on a shop floor does not. Photographing a garment on the rack, hung straight and shot square-on, is enough.

When it saves you money

Try-on earns its place in a few specific situations rather than every purchase:

• Online orders you would otherwise return. Return shipping is increasingly charged to the buyer, and a flat-lay product photo tells you nothing about how a cut sits on your frame.
• Second-hand and vintage. One of each item, no returns, and usually a photo taken on somebody's floor. The stakes for guessing wrong are highest here.
• Two options, one budget. Trying both in a shop means undressing twice and still comparing from memory. Two try-on images sit side by side.
• Anything you're buying on impulse. Twenty seconds is short enough that you'll actually check, and long enough to break the impulse.

Privacy: where your photo goes

You are handing an app a photograph of yourself, so it is worth asking three questions before you install anything:

• Does it need an account? If a try-on app requires sign-up, ask what the profile is for. The generation itself doesn't need one.
• Where are the images stored? On your device, or in the company's cloud indefinitely?
• Is your photo used for training? Check the policy, not the marketing page.

Hoilo's own answers: no account, images stored in the app on your phone, and your photo sent to the image model only to produce the try-on you asked for. The full text is in the privacy policy.

Trying a garment on right now

Hoilo is an AI virtual try-on app for iPhone. You give it a garment — photograph one in a shop, paste a product link from any online store, or share a screenshot into it from any app — and one photo of yourself. It returns you wearing the item in about 20 seconds, grouped in a history by product so five attempts at the same coat stay together.