ROLL Beauty

Inspiration

Describing a haircut to someone — or remembering what your own hair looked like last month — is surprisingly hard. A single photo only shows one angle. We wanted something closer to actually turning around in front of a mirror.

What it does

ROLL Beauty records a 360° "turntable" of your hairstyle: point the front camera at yourself, turn around once, and the app captures 36 frames, aligns and background-composites them, and lets you view the result by dragging — like spinning a physical object. Stylists can use it to show off their work; anyone can use it to see their own hair from every angle.

How we built it

• Capture: ARKit face tracking measures rotation angle while the face is visible, then falls back to elapsed-time estimation past profile.
• Processing: Vision's person segmentation runs in two passes across all 36 frames — first to find the median face width, then to segment hair and shoulders consistently — and Core Image composites a clean white background.
• Backend: Supabase (Auth, Postgres, Realtime) powers accounts, a lightweight social layer, and stylist verification.
• Monetization: RevenueCat + RevenueCatUI power the subscription paywall.

Challenges we ran into

Getting the rotation to feel natural took real-world testing — actual turn speed varied 3–14 seconds across takers, so we drive the pacing with an audio cue instead of guessing. Keeping the hairline crop stable across 36 independently-segmented frames also took a two-pass measurement approach rather than per-frame guesses.

Accomplishments that we're proud of

Shipped to the App Store as a two-person team, with a working ARKit + Vision capture pipeline that runs entirely on-device.

What's next

Expanding the social/sharing side, and iterating on hairline detection accuracy from real salon usage.

Built With

• arkit
• core-image
• postgresql
• revenuecat
• supabase
• swift
• swiftui
• vision