Vocaa
In production
<2s
Match time
5 min
Session
15+
Languages
Real-time language speaking practice: one button, instant match with a human partner, five-minute voice sessions. Designed for immigrants learning German, global English learners, and anyone who wants accountable speaking reps — without scheduling friction or fake AI small talk.
The stack mirrors the ambition: React Native (Expo) for native mobile, NestJS services, PostgreSQL on Supabase, Redis via Upstash for fast session state, LiveKit for WebRTC, Whisper + GPT-4o for pronunciation intelligence where solo feedback helps. Railway hosts the backend; Vercel ships the web surface; EAS builds Android. Features include sub-two-second matching, streaks, timed sessions, Quick Pronunciation Check, and support for fifteen-plus languages.