Back to catalogue
Project

Smart Buddy AI
Interactive Gemini tutor prototype for junior secondary school learners.
How it’s builtHow it’s built
Smart Buddy AI · Real-time voice tutor for junior secondary students on Gemini Live
I built a browser voice tutor for junior secondary school learners that streams microphone audio at 16 kHz to the Gemini Live native-audio model and plays back 24 kHz spoken replies with live transcripts. Learners pick one of 6 JSS subjects with suggested topics, and a curriculum-grounded system prompt shapes the answers.
- JSS subjects
- 6
- input / output audio sample rates
- 16 kHz / 24 kHz
- lines of TS/TSX
- 464
Technology and what it does here
- @google/genai Live API
- Ai.live.connect with gemini-2.5-flash-native-audio-preview, AUDIO modality and Kore voice
- Web Audio API
- ScriptProcessor captures mic PCM; decodeAudioData plays streamed replies
- React 19 + Vite
- Subject picker, topic chips, live status and transcript UI
- TypeScript
- Typed subjects, messages and transcription entries
Implementation details
- Microphone frames are sent as 'audio/pcm;rate=16000' blobs and model audio parts are decoded from base64 inlineData into an AudioBuffer.
- Output transcription is accumulated per turn and pushed into the chat history when the turn completes.
Technologies
- React
- Vite
- Gemini
