People buy beauty products blind. They guess their shade, guess what their skin needs, and return a big share of what they bought. GlowCart replaces the guessing with measurement — one selfie, three YouCam APIs, one continuous journey from diagnosis to purchase.
Scan — YouCam AI Skin Analysis. One front-facing selfie is analyzed for 12 skin concerns (moisture, wrinkles, acne, pores, texture, oiliness, redness, dark circles, eye bags, firmness, radiance, age spots), an overall skin score and an estimated skin age. GlowCart ranks the concerns weakest-first, in plain language.
Advise — GlowAdvisor. A deterministic reasoning layer reads the actual numbers and explains the plan: what your skin needs first, why each recommended product earned its place (tied to its score), what your budget should protect versus skip, and an AM/PM routine built from the picks. Change the budget filter and the advice re-derives itself live.
Match — YouCam AI Skin Tone Analysis. The same selfie goes through skin tone analysis. GlowCart converts the measured skin colour into CIELAB space, classifies depth with the ITA° method used in dermatology, derives undertone from the a*/b* balance, and ranks a 12-shade foundation catalogue by colour distance (ΔE) with an undertone-mismatch penalty. The top three matches each come with the reason they fit.
Try it on — YouCam Makeup VTO. Pick a matched shade and the Makeup VTO API applies that exact foundation to your own photo. A before/after slider proves the match on your face — before the product ever reaches your cart.
Buy. The tried-on foundation joins the skincare routine in a single cart, trimmed to your budget (under $50 / $100 / $200 / no limit), with anything the budget held back shown honestly, and a checkout summary with an AM/PM timeline.
Three APIs working as one product — not three demos stitched together: the scan feeds the advisor, the tone feeds the match, the match feeds the try-on, and the try-on feeds the cart.
Live-API note: The deployed app runs on the real YouCam APIs — verified end to end with a real user photo: Skin Analysis returned real measured scores (overall 82), Skin Tone returned a measured colour, and Makeup VTO returned a real rendered try-on image (server-side YouCam V2 client: file upload → task → poll, per feature). Honest by design: if a photo cannot be scanned, GlowCart never invents results — it offers a self-guided path where the user picks their own concerns, clearly labeled, with real products matched to each pick. A clearly labeled demo fallback remains in the code for key-less environments, using realistic sample payloads in the documented response shapes through the same normalizers.
Built With
- ai
- analysis
- api
- css
- makeup
- next.js
- skin
- tailwind
- tone
- typescript
- vercel
- vto
- youcam
