Cinemi

A mobile-first AI image studio driven by photography slash-commands. Pick a photo, type /85mm or /filmgrain, compare before and after, retouch faces on-device.

Sole engineer & designer · 2026 · Designed, built and deployed by Alexus Karl Joaquin.

What I built

• "Editorial After Dark" design system: primitives, semantic tokens, dark + light themes, motion and safe-area tokens. The reference page reads token values live from the CSS, so documentation cannot drift from code.
• On-device face retouching with MediaPipe's face landmarker running in WebAssembly. Photos never leave the phone.
• Slash-command prompt composer with a searchable command catalogue, prompt preview and compile step; before/after compare slider; image viewer with favorite and remix.
• Monetization components designed in: credit meter, rewarded-ad offer and confirmation, labelled ad slots that reserve their size and never sit near creative controls.
• Built for Capacitor from day one: self-hosted variable fonts (works offline), keyboard-inset handling, 44px touch targets, hover only on fine pointers.

Skills demonstrated

• React 19
• TypeScript 6
• Vite 8
• Tailwind CSS 4
• Radix UI primitives
• shadcn/ui conventions
• class-variance-authority
• Design tokens (primitive to semantic)
• MediaPipe on-device ML (WASM)
• Canvas image processing
• Self-hosted variable fonts
• Mobile-first UX for Capacitor
• Accessibility (touch targets, reduced motion)
• Monetisation UX (credits, rewarded ads)
• ESLint + strict typecheck tooling

Stack

Screens

More

All projects · Experience · LinkedIn · GitHub