Adding Voice to a Java AI Assistant — Whisper, TTS, and the Voice Conversation Loop

Short summary

Jarvis Phase 5 adds voice by routing Whisper transcription and TTS through the existing chat pipeline, avoiding a separate voice stack. Using Groq's free API or whisper.cpp for transcription and native OS engines for speech synthesis keeps dependencies minimal. Key techniques include sentence buffering for natural playback and Spring WebFlux's boundedElastic() scheduler to prevent blocking I/O on the event loop.

• •Voice wraps existing text pipeline via Whisper + TTS, no separate architecture
• •Groq API (free tier) or local whisper.cpp for transcription; native OS engines for synthesis
• •Sentence buffering and boundedElastic() scheduler solve streaming and concurrency challenges

Generated with AI, which can make mistakes.