Silero voice detection --> Whisper STT --> Claude Opus 4.5 API call --> Kyutai TTS equals responsive, humanlike discord Clawdbot voice call! Responses take like 4-6 seconds because of the Opus response generation, but otherwise running pretty fast on my local GPU!

Honestly I was thinking the same thing

Same thing, asked Claude to make a YTP about “what it means to be an LLM”