Hermes · Voice Mode

Hold a key, talk to Hermes, hear it talk back. CLI, Telegram, Discord, and Discord VC.

Voice mode lets you talk to Hermes on the CLI (push-to-talk with Ctrl+B, VAD silence detection), on Telegram and Discord (auto voice replies alongside text), and in Discord voice channels (the bot joins, listens, speaks back). Three STT picks: faster-whisper (local, no key, default), Groq Whisper (free tier, fast), OpenAI Whisper (paid). Four TTS picks: Tool Gateway (Portal, no keys), ElevenLabs (premium), Edge TTS (free), NeuTTS (local). New since the last episode: live transcription, set stt.streaming: true (OpenAI, xAI, or ElevenLabs) and partial text appears while you speak, with automatic fallback to file transcription. On Telegram and Discord, replies are slash-command modes (/voice, /voice tts, /voice off), persisted across gateway restarts. The stop word must be the whole utterance, and hands-free session start is the separate wake word feature.