Maison Osùn

Inspiration

Osùn is camwood, a red heartwood that families across West Africa have pressed onto their skin for generations. Good skincare used to be personal: someone looked closely at your skin and handed you exactly what you needed.

Online, that care gets lost in a shelf of 40 serums and a quiz. I wanted to bring back the luxury beauty counter, where someone reads your skin and hands you three things, for anyone with a phone.

What it does

Maison Osùn is a full app, not just a landing page:

• Consultation: YouCam's JS Camera Kit guides you into good light, distance and angle, then takes the selfie on its own. Plain selfie or upload also works.
• Portrait: YouCam Skin AI scores seven signs (hydration, oil balance, radiance, texture, pores, redness and blemishes), plus an overall score, skin age and skin type. Tap a sign to see YouCam's detection mask on your own face. A Clinical view shows the raw scores, and every reading shows its YouCam task ID.
• Day 28 preview: YouCam AI Skin Simulation shows your own skin after the ritual, targeting your two weakest signs, as a before/after slider. You see it before you buy anything.
• Ritual: a morning and evening routine built from your scores, where every product says which score it's for. A daily tracker lets you tick off steps, keep a streak, and count down to your next portrait.
• Journal: every reading is saved on your device, with a chart for any sign over time, so you can see whether the ritual is actually working.
• Bag: the whole ritual in one tap.

How I built it

• YouCam Skin AI (V2 API): POST /file/skin-analysis for an upload slot, then a PUT to the pre-signed URL. Next, POST /task/skin-analysis with seven concerns, and polling GET /task/skin-analysis/{id} for the scores and masks.
• YouCam AI Skin Simulation (V2 API): the two weakest signs become intensities from 0.2 to 0.6, so weaker skin gets a stronger but still natural change. These go to POST /task/skin-simulation, which is then polled for the result image.
• YouCam JS Camera Kit in skincare mode for guided capture.
• Next.js 14 on Vercel. Every API call goes through server routes, so the API key never reaches the browser.
• Readings, ritual check-offs and the bag are stored on the device, so no photos or scores sit on a server.

Challenges I ran into

• Getting a usable photo. The reader needs the face to fill most of the frame. I resize photos in the browser to the API's limits, and turn every YouCam error code into one plain instruction, like "Come closer" or "Find softer light".
• Making scores feel like advice, not a report card. Every product explains in one line which score it answers.
• Trust. YouCam's display scores run flattering, so I added the Clinical view with raw scores and showed the task ID on every live reading.

Accomplishments that I'm proud of

• Using three YouCam tools together (Skin AI, AI Skin Simulation and Camera Kit) in one coherent product journey.
• The day-28 preview: seeing your own face improve before buying is the moment that sells the ritual.
• A design that feels like a luxury house, not a hackathon template.

What I learned

• How YouCam's task-based V2 API works end to end: upload slots, async tasks, polling, and the difference between ui_score and raw_score.
• That AI results need explaining to feel trustworthy: provenance and plain-language reasons matter as much as the scores.

What's next for Maison Osùn

• Accounts so the journal syncs across devices.
• Real checkout, plus a 28-day refill subscription.
• Re-composing the ritual automatically when a new reading shows a sign has recovered.

Built With

• framer-motion
• lenis
• next.js
• react
• youcam-ai-skin-simulation-api
• youcam-js-camera-kit
• youcam-skin-ai-api