Voice Tutor

TutorQ's voice tutor provides real-time speech-to-speech tutoring powered by AWS Nova Sonic. Students speak naturally, and the AI responds with voice — grounded in the course materials.

How It Works

Student speaks a question
    ↓
Audio streamed to TutorQ (16kHz PCM)
    ↓
Nova Sonic processes speech and decides whether to search course materials
    ↓
If needed: RAG search finds relevant passages from uploaded materials
    ↓
Nova Sonic generates a spoken response using the found content
    ↓
Audio streamed back to student (24kHz)

Key Features

Curriculum-Grounded

Every answer comes from the professor's uploaded materials. The AI uses a search_course_materials tool to find relevant passages before responding. No hallucinations from generic training data.

Voice-First

Students speak naturally — no typing. The AI responds with voice. This is especially valuable for:

• Students who struggle with written text
• Mobile users
• Hands-free study sessions
• Students with reading difficulties

Adaptive Teaching

Six discussion modes that adapt to how the student wants to learn — from direct explanation to Socratic dialogue to guided reading.

Barge-In Support

Students can interrupt the AI mid-response (barge-in), just like in a real conversation. The AI stops immediately and listens.

Multi-Language

Voice selection adapts to language. Currently supported: English (matthew, tiffany, gregory, stephen voices) and Hindi.

Technical Details

Session Flow

1. Connect — WebSocket connection established
2. Init — Client sends init message with course ID and language
3. Session start — Nova Sonic session created with system prompt and course context
4. Audio loop — Bidirectional audio streaming (student ↔ AI)
5. Tool calls — AI searches course materials when relevant
6. End — Client sends end_session, metrics saved to database