Virtual Try-On for Beauty Marketplaces: One Integration, Many Brand Palettes

A marketplace is not a brand site with more SKUs. It is dozens or hundreds of sellers, each with its own shade naming, its own photography, its own idea of what "nude" means. That is exactly why virtual try-on for beauty marketplaces is a different engineering and governance problem from putting a try-on button on a single brand's product page: the shopper experience has to feel like one feature, while the colour data underneath belongs to many different catalogs.

This article covers how to structure that — the data you need from sellers, where try-on actually belongs in a marketplace journey, and how to roll it out category by category without blocking launch on your longest tail of sellers.

Why shade data breaks first

On a single-brand site, the colour team can review every lipstick and sign off on the palette once. On a marketplace, colour information arrives through a seller portal, a spreadsheet or a feed, and it arrives in whatever shape the seller sends it.

Typical symptoms:

• The same shade name appears across three brands with three different pigments.
• Foundation shades carry a code ("120 W") but no reference to tone or undertone.
• Product images are the only colour source, and they were shot under different lighting.
• Swatch crops are inconsistent — some on skin, some on white, some on the packaging.

If try-on is wired straight to that feed, the preview inherits every inconsistency. The fix is not to clean the entire catalog before launch. It is to decide which sellers and which categories get a calibrated palette first, and to treat palette calibration as an onboarding step with an owner, not as a background data job.

Virtual try-on for beauty marketplaces: how the pieces fit

There are two delivery shapes, and marketplaces usually end up using both for different surfaces.

Photo-based rendering through the API

With MaIA Try-On as an API, the shopper's photo is sent in, the product is rendered on it, and the image comes back. The API covers lipstick and foundation, and palettes are calibrated to each brand's real products inside a project. For a marketplace, that project boundary is useful: seller A's lipstick line and seller B's lipstick line are separate palettes, maintained separately, called through the same integration.

Because the output is an image, it travels. The same call can serve a product page, a WhatsApp conversation, a CRM email asset or a campaign landing page — anywhere you can display a returned image rather than run a camera.

Live camera mode through the browser SDK

The MaIA Try-On browser SDK, in early access since September 2026, is a different tool for a different moment. It covers thirteen products — lipstick, foundation, blush, face powder, under-eye concealer, eyeshadow, mascara, eyebrow makeup, eyeliner, bronzer, contour, highlighter, lip liner — plus a combined Full Look, in Live mode using the shopper's camera rendered locally in the browser, or in Photo mode. It accepts any #RRGGBB colour and is self-hosted by the marketplace.

Two consequences matter for a marketplace. First, Live mode keeps the frames in the browser — nothing is uploaded — which simplifies the privacy conversation with legal and with sellers. Second, because the SDK takes hex values, a seller can be live with a plausible preview as soon as it provides reliable hex codes, while deeper calibration happens on the API side for the categories that need it most. The SDK is granted per account, so plan the request alongside your roadmap rather than assuming a public download.

If you are still deciding which surface deserves which approach, the comparison in photo-based try-on vs AR try-on lays out the trade-offs without the marketing vocabulary.

What to require from sellers

Treat try-on readiness as a tier in your seller onboarding, not a universal requirement.

Minimum viable colour record

For each shade, ask for: a stable SKU, the shade name as printed, a hex value, the finish (matte, satin, gloss, metallic) and the category. Finish matters because the same hex reads differently as a matte lipstick and a gloss. Without finish, previews look flat and sellers complain that the colour "is not theirs".

Ownership of the palette

Decide who signs off: the seller, your category manager, or both. In practice a short approval loop works — the marketplace produces the first palette, the seller reviews a rendered sample on two or three different skin tones, and approval is recorded against the SKU. When a seller later changes a formula, the sign-off is the trigger to recalibrate.

A visible fallback

SKUs without an approved palette simply do not show the try-on entry point. A missing button is invisible; a wrong colour is a support ticket and a trust problem for the whole marketplace.

Where try-on belongs in a marketplace journey

• Category and search results. A "preview this shade" affordance on the grid reduces the number of dead-end product page visits, but only for categories where colour is the main decision.
• Product page. The default home for try-on, next to the shade selector, with the seller's name visible so the shopper knows whose colour they are seeing.
• Cross-seller comparison. The most marketplace-native use: three lipsticks from three sellers previewed on the same photo. This is where a marketplace can do something a single brand cannot.
• Post-purchase and CRM. Photo-based renders in lifecycle messages work well for replenishment and for pushing adjacent shades.

Pairing try-on with a beauty profile sharpens the shortlist before the preview: MaIA Scan returns skin tone and undertone, skin type and condition, hair attributes and a confidence per field from a selfie, which lets you order the shade grid by relevance instead of by seller ranking. It describes visible attributes in a photo — it does not diagnose anything.

Measuring it honestly

Measure by cohort, not in aggregate: shoppers who previewed versus comparable shoppers who did not, split by category and by seller tier. Watch add-to-cart on previewed SKUs, shade changes before purchase, and return reasons that mention colour. Published industry figures from other companies are not your forecast — results vary with each catalog, traffic mix and channel, and a marketplace's mix changes every month.

Ready to test the rendering against your own shade feed? Create an account and get an API key, or talk to the B4A team about palette calibration for your top sellers.