WinnerSystems
Voice product · HackPrinceton Winner
Saynario
Won HackPrinceton Best Use of ElevenLabs with a shipped voice roleplay loop.

Overview
Saynario is a voice-first language learning app built around live roleplay rather than static drills. Each conversation feeds the next one, so practice stays adaptive instead of resetting back to generic prompts.
Problem
Most language-learning tools make it easier to review vocabulary than to build comfort speaking spontaneously. The harder problem is turning live conversation into useful, personalized follow-up practice.
Approach
- Built a real-time roleplay loop across the React client, FastAPI backend, WebSockets, speech transcription, and text-to-speech output.
- Used Gemini-powered adaptive coaching so each finished conversation could generate the next personalized scenario.
- Deployed the system end to end so the product felt like a usable voice interface rather than a disconnected model demo.
Result
Won HackPrinceton 2026 Best Use of ElevenLabs and produced a working voice practice experience centered on real conversation instead of passive repetition.
Stack
ReactFastAPIWebSocketsGeminiClaudeWhisperElevenLabsDigitalOcean