Storis
Overview & architecture
Problem & scope
Sending live voice audio to cloud speech APIs during bedtime reading introduces noticeable network latency, drains battery, and forces families to stream microphone streams off-device. Audio cues needed to trigger accurately against spoken phrases in real time right on the phone while preserving calm, customizable reading screens.
System architecture
Pairs low-overhead microphone capture via record with local speech-cue inference inside the Sherpa ONNX runtime, routing detected triggers straight into audioplayers soundscape channels without cloud speech round-trips. Riverpod coordinates library state across Supabase cloud storage and Apple/Google Sign-In, backed by RevenueCat entitlements, Envied build-time secrets, and SharedPreferences caching.
Key engineering highlights
- Executes edge speech-cue detection locally via Sherpa ONNX for minimal audio lag and on-device voice privacy.
- Synchronizes spoken narration with dynamic soundscape playback alongside light/dark reading themes and adjustable typography.
- Delivers full production SaaS infrastructure including OAuth authentication, iOS subscription management, local state persistence, and Envied secret handling.
Technologies used
- Flutter
- Sherpa ONNX
- Supabase
- Riverpod
- RevenueCat
Home screen of Storis
More projects
Platterify
Platterify is a real-time multiplayer hospitality platform for restaurants and bars. Guests scan table QR codes to browse digital menus, order collaboratively, and request bills in one tap, while staff manage live tables and kitchen tickets.
Kippr
Kippr is an offline-first pantry tracker and inventory mobile application. It helps you easily manage your kitchen stock, reduce food waste, save money, and organize your groceries right from your phone.
Knowla - Web
Knowla - Web is an interactive platform that supports children with disabilities through game-based learning. Using multimedia and gamification, it boosts cognitive skills, creativity, and problem-solving.