Fluenta – The Voice Tutor That Waits for You

Learning a language means pausing to find the right word. Most voice agents are tuned for fluent speakers, so they treat that pause as the end of your turn and interrupt you right before you get it right. Fluenta is built around fixing that. Fluenta is a voice-first language tutor for Spanish, English, French and German. Sign-up is done entirely by voice: Sabio, our owl tutor, asks which language you speak, what you want to learn and your level. Lessons are role-play scenarios with missions, like ordering at a café, inside a timed class. Key features: - Level-aware patience: turn detection adapts to the learner. A1 students get up to 1.2 s of silence before their turn ends; B1 students get 0.7 s. - Help in your language: one tap switches Sabio to the learner's native language with a native voice, then returns to the lesson with the conversation carried over. - Live coaching: new vocabulary and gentle corrections appear as cards while you speak. - Honest reports: an end-of-class report with points, mission score and feedback. Our server discards any correction that doesn't quote what the learner actually said, so the model can't invent mistakes. How it's built: the browser streams PCM16 audio directly to the AssemblyAI Voice Agent API using single-use tokens. A small Node server issues tokens, enforces per-IP and daily session limits, and generates coaching and reports through the AssemblyAI LLM Gateway with strict JSON schemas. The API key never reaches the browser. Frontend: React, Vite and Tailwind.