Guide · virtual makeup try-on
Virtual try-on vs AR try-on: what's the difference, and which one should your brand choose?
"Virtual try-on", "AR try-on", "virtual mirror", "photo-based try-on" — the terms get mixed up, but they describe two different technologies with different costs, channels and outcomes. This guide explains each one without the marketing, shows where each approach works best and helps you decide which makes sense for your e-commerce, your WhatsApp or your store.
What virtual try-on is
Virtual try-on is any technology that shows a shopper how a beauty product — lipstick, foundation, eyeshadow, hair colour — looks on her before she buys. It is the umbrella term. Under it sit two technical families: live augmented reality (AR try-on), which applies the product over the camera feed in real time, and photo-based virtual try-on, which takes a still image and returns the same image rendered with the product.
The confusion comes from history: the first generation of virtual try-ons, roughly 2017 to 2021, was almost entirely AR — magic mirrors in stores and camera filters in brand apps. Today, with generative models trained specifically for makeup, photo-based try-on delivers a more faithful result in many cases — and fits channels AR never reached, such as WhatsApp and e-mail.
What AR try-on (live augmented reality) is
In an AR try-on, an SDK on the site or in the app accesses the camera, detects facial landmarks frame by frame and draws the product over the video in real time. The person turns her head, smiles, the light changes — and the lipstick follows. It is a mirror experience: immediate, playful, great for exploring dozens of shades in a row.
The price of that immediacy is infrastructure: WebGL or native code on the client, camera permission, a reasonable device and a stable connection. The final colour depends on the AR engine and the camera lighting at that moment, which makes fidelity to the real product hard to guarantee. And by default nothing remains: the experience ends when the camera closes, unless the app records a capture.
What photo-based virtual try-on is
In a photo-based try-on the customer sends a front-facing image — an upload, a photo taken on the spot or a WhatsApp message — and a computer-vision model locates the makeup area and renders the product with the right colour, coverage and finish. The result comes back as an image file within seconds.
B4A's MaIA Try-On works this way: one HTTP call with the photo, the shade and the finish, and the photo returns as a PNG in about 2 to 4 seconds, with only the makeup area changed — teeth, mouth, skin and features untouched, no smoothing, no reshaping. Lipstick is the launch category (six demo shades in matte, satin and gloss); other makeup categories and a palette calibrated to the brand's products are delivered as a project. The photo is not stored.
Comparison: photo-based try-on vs live AR
| Photo-based virtual try-on (e.g. MaIA Try-On) | AR try-on (live camera) | |
|---|---|---|
| Input | One still, front-facing photo | Real-time camera video |
| Where it works | Any channel that carries an image: PDP, quiz, WhatsApp, e-mail, kiosk | Site or app with an SDK, WebGL and camera permission |
| Integration | One HTTP call; no SDK, no app | Front-end SDK, often one per platform |
| Latency | Seconds per render (2 to 4 s in MaIA Try-On) | Instant, frame by frame |
| Output | A file the customer saves, compares and shares | A live experience; a capture only if the app records one |
| Colour fidelity | Calibrated per product, with human validation | Depends on the AR engine and the camera light |
| Device requirements | None beyond taking or sending a photo | Camera, a reasonable GPU, a stable connection |
| Best for | E-commerce, conversation, quizzes, e-mail, repurchase | Store mirrors, in-app experiences, playful exploration |
When photo-based try-on is the right choice
- The sale happens on the product page: a "see it on me" button next to the shade selector, with no SDK and no camera prompt, answers the shade question right where it appears.
- The channel is conversational: on WhatsApp the customer sends a selfie and gets her photo back with the product and the link — no AR mirror fits inside a conversation.
- You want a result that lasts: the file goes to the gallery, to the group chat with friends and into the abandoned-cart e-mail.
- Fidelity to the product matters more than play: every shade is calibrated against the real lipstick, on several skin tones, before going live.
- You need to integrate fast and on several platforms at once: an HTTP API works in any stack — web, app, CRM, messaging platform.
When AR try-on is the right choice
- A mirror in a physical store or at an event: the live experience, with movement, is what attracts and entertains.
- Exploration inside the brand's app: switching fifteen shades in a row, in real time, is smoother in AR.
- Categories that need movement to convince: the shine and reflection of a gloss or a highlighter under changing light.
- You already have a native app with camera access and a front-end team to maintain one SDK per platform.
The right questions before you decide
Where is your customer when she decides? If it is on the PDP, on WhatsApp or in e-mail, photo-based try-on slots in without friction. If it is inside your app or in the store, AR may justify the SDK investment.
What do you want to remain after the experience? A shareable file drives reach and cart recovery; a mirror drives dwell time and delight.
How faithful does the colour have to be? For lipstick, foundation and hair colour a wrong shade becomes a return. Per-product calibration weighs in favour of photo.
How much can you integrate? One HTTP call versus one SDK per platform changes the project's timeline and cost — and who maintains it.
They can coexist
The two approaches do not compete for the same spot. A brand can run an AR mirror in its concept store and photo-based try-on on the PDP, on WhatsApp and in the quiz — and because the customer profile is the same, MaIA Beauty Advisor can use the photo try-on result to recommend the next product in the routine. What matters is choosing the technology by channel and purchase decision, not by whichever term is trending.
If you want to see photo-based try-on at work, the MaIA Try-On demo takes your own photo: six sample shades, three finishes, a result in seconds and the photo discarded right after.
Update (September 2026): MaIA Try-On now also offers a live camera mode, with thirteen products and a Full Look, through a browser SDK in early access — rendered on the customer's device, without uploading the photo. The photo API remains the SDK-free path.
See photo-based try-on on your own photo.
The MaIA Try-On demo is on the home page: six sample shades, three finishes, a result in 2 to 4 seconds. Or talk to us about your brand's palette.