Inspiration
I grew up surrounded by pregnant women: aunts, cousins, friends. I watched my aunts try every home remedy on their faces, from aloe vera to breast milk, always with the best intentions and almost always blindly.
No two pregnancies are alike, and no two skins are either. Some friends glowed, then fought intense dryness postpartum; others battled acne and rosacea that slowly faded. Every story was different, but they all shared worry, uncertainty and very little guidance. And some common skincare ingredients can harm the baby, yet many women don't know it.
That's how Ninemonths was born: AI that analyzes your skin, catches what is changing in time, and recommends what is safe for you and your baby.
What it does
- Weekly guided selfie. On-device checks (face, distance, angle, light) and last week's photo as a ghost guide make every week comparable.
- YouCam Skin Analysis: 6 pregnancy-relevant concerns (dark spots, acne, redness, hydration, radiance, dark circles).
- Melasma, zone by zone. YouCam's spot mask is mapped onto 5 face zones and tracked by pregnancy week: "Your melasma is growing on your right cheek." A measured noise floor means it only reports real changes.
- Tinted sunscreen try-on (YouCam Makeup VTO). Tinted mineral SPF is the best melasma defense: see 3 shades around your measured tone on your own face, with no airbrushing.
- "Is this cream safe for my baby?" Photograph the label; every ingredient gets a pregnancy traffic light, a medical source (ACOG, AAD) and a safer alternative.
- Symptom check with red flags (itchy palms → possible cholestasis), a routine for your week and local UV, and a one-page summary for your doctor.
- Private: no account; your history stays on your phone.
- Business model: free tier + a $9.99 one-time Pregnancy Pass.
Try it in one minute, no camera needed: ninemonths.skin/demo
How we built it
Next.js 16, React 19, TypeScript and Tailwind on Vercel. Server-side YouCam Skin Analysis v2.1, Fitzpatrick Scale and Makeup VTO, with tasks deleted after reading. MediaPipe Face Landmarker on the device for capture checks, the YouCam-compliant crop (rejected photos cost no units) and the melasma zones. On-device OCR (tesseract.js) and a sourced pregnancy ingredient rule set. 35 automated tests, including a guard that fails the build if any text promises a diagnosis or a cure.
Challenges we ran into
- Making weeks comparable. Light and pose move the scores, so we measured the noise (6 captures of the same face) and only report changes above it.
- False spots. Our first zone map counted hair, lips and side light; per-zone lighting correction and masks fixed it, verified with controls.
- API details: dark circles need
dark_circle_v2; VTOskin_smoothneeds both strength and color intensity.
Accomplishments that we're proud of
- Live end to end at ninemonths.skin on the real YouCam APIs (about 6–10 s per analysis).
- Skin AI + VTO in one journey: measure the melasma, then try the protection on your own face.
- Same face, uploaded vs. guided camera: zone values within 1 point.
What we learned
Raw scores mean little until you know their noise. And an anxious mom-to-be wants a clear next step with a source behind it.
What's next for Ninemonths
Postpartum and breastfeeding mode, validation with OB-GYNs and dermatologists, a pilot with prenatal clinics in Latin America, and shade-matched tinted sunscreen shopping.
Built With
- indexeddb
- mediapipe
- next.js
- open-meteo
- react
- tailwindcss
- tesseract.js
- typescript
- vercel
- vitest
- youcam-fitzpatrick
- youcam-makeup-vto
- youcam-skin-analysis