Sephora
virtual AR try-on (synthetic makeup rendering on the face)
Sephora Virtual Artist, the AR try-on built with ModiFace, has let customers try on hundreds of millions of shade combinations since 2016 (more than 200 million shades and 8.5 million visits cumulatively by 2018).
Key points
- Virtual AR makeup try-on via the camera (app, web, in-store kiosk).
- ModiFace engine: facial mapping, skin tone analysis, and real-time rendering on a catalog of calibrated shades.
- Hundreds of millions of shades tried on since 2016 (200M shades, 8.5M visits by 2018).
- Evidence level B, confirmed status.
Objective
Have thousands of shades tried on virtually to remove color doubt, increase conversion, and reduce returns on makeup.
The deployment
Sephora Virtual Artist, launched in 2016 and built with ModiFace, lets customers virtually try on thousands of makeup products from the app or the web via the camera. The system does facial mapping and skin tone analysis to apply a realistic rendering (lips, eyes, cheeks), offers expert looks and tutorials, and a before/after split-screen view. The feature is also available in store. Sephora operates more than 2,300 stores in 33 countries and 14,000 products from 200 brands.
The case in action
Official videoSephora Virtual Artist : essayage AR · voir sur YouTube
Results Proof B
Usage figure cited in an official Sephora release (primary source, VP Innovation) and confirmed by the press. Multiple aligned sources on the same feature.
How it works
Documented architectureThe stack in detail
- outil ModiFace AR try-on engine: facial mapping, skin tone analysis, and real-time makeup rendering (lips, eyes, cheeks)
- infra Application Sephora / SephoraVirtualArtist.com in-house try-on surfaces: mobile app, web, and in-store kiosks
- infra Catalogue de teintes calibrees product reference set (14,000 items, 200 brands) whose shades are calibrated so the AR rendering matches the real product
How it runs, concretely
For ops teams-
1Select a product or look customer
The user picks a shade, an expert look, or a tutorial in the app.
-
2Facial mapping AI (ModiFace)
The engine maps the face and analyzes skin tone via the camera.
-
3Rendering and comparison AI (ModiFace)
The product is applied in real time; the split screen shows before/after.
-
4Purchase customer
The user adds the virtually validated items to the cart.
-
5AR catalog update Sephora / ModiFace team
Sephora and ModiFace add new items and recalibrate the renderings.
The match between the product's real shade and the AR rendering. A poorly calibrated shade distorts the try-on and can generate returns instead of reducing them.
How your customers perceive this type of use
Sourced studiesUn ecart net separe les annonceurs des consommateurs : 77% des annonceurs voient l'IA positivement contre 38% des consommateurs (Yahoo/Publicis, 2024). Les mesures implicites confirment le rejet declare : en EEG, les pubs generees par IA produisent une activation memorielle plus faible que les pubs traditionnelles et sont decrites comme agacantes, ennuyeuses et confuses (NIQ, 2024). La disclosure a un effet ambivalent : elle augmente fortement la confiance quand elle est remarquee (Yahoo/Publicis), mais 27% des jeunes consommateurs disent faire moins confiance a une entreprise dont la pub est creee par IA (IAB, 2024).
Acceptance conditions
- Une disclosure visible : quand la mention IA est remarquee, la confiance globale envers l'entreprise augmente de 96% (Yahoo/Publicis 2024)
- Une qualite visuelle suffisante : les visuels IA de basse qualite augmentent l'effort cognitif et distraient du message (NIQ 2024)
Red lines
- Le contenu IA non declare puis identifie : 72% des consommateurs disent que l'IA rend l'authenticite difficile a etablir (Yahoo/Publicis 2024) et les marques utilisant des pubs IA sont plus souvent jugees inauthentiques ou non ethiques par les consommateurs que par les dirigeants (IAB 2024)
- Les mannequins et personnes generes par IA : 46% des consommateurs n'en veulent pas dans la publicite, l'inquietude premiere etant les standards de beaute irrealistes (Attest 2025)
Sources: Yahoo / Publicis Media (terrain Ebco) 2024 · IAB (avec Attest) 2024 · NIQ (NielsenIQ) 2024 · Attest 2025
How to replicate
Inference, not sourcedData prerequisites
- makeup product reference set with calibrated shades
- visual assets of the looks
Org prerequisites
- app/e-commerce integration
- continuous calibration process
Possible stack
- ModiFace
- Perfect Corp (YouCam)
- Banuba
The plan, step by step
- Step 1Choose the AR provider (ModiFace, Perfect Corp, Banuba) on rendering realism and product coverageDeliverable: Provider selected and contract signed
- Step 2Calibrate the best-seller shades: match the real shade and the AR rendering on the highest-traffic itemsDeliverable: AR catalog calibrated on the top sellers
- Step 3Integrate the SDK into makeup product pages, with explicit camera consent and non-retention of face images (GDPR)Deliverable: Live try-on on a pilot scope
- Step 4Track shades tried on, conversion, and returns against pages without try-on, then extend to the other categoriesDeliverable: Quantified readout, extension plan, and continuous calibration process for new items
First step: Enable AR try-on on the highest-traffic makeup categories and calibrate the best-seller shades.
The same pattern, proven in other industries
Leroy Merlin (Adeo)
genAI enrichment of product attributes and pages
eBay
GenAI generation of a product listing from a photo
Intuit
done-for-you genAI assistant integrated into the product to reduce friction and convert
Sources
- S1 Sephora Virtual Artist Adds Virtual Try On Of Thousands Of Eyeshadow Shades, New Expert Looks And An Expanded Library Of Virtual Tutorials Primary archive pending
- S2 Sephora's Virtual Artist brings augmented reality to large beauty audience Established press archive pending
- S3 How Sephora is leveraging AR and AI to transform retail and help customers buy cosmetics Secondary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.