Skip to content

VOICE.md

Latest commit

 

History

History
42 lines (26 loc) · 1.52 KB

VOICE.md

File metadata and controls

42 lines (26 loc) · 1.52 KB
Raw

Voice stack

Modes (user-facing)

Mode Flags Input Output
/voice voice_mode_active=True, agent_voice_active=True Mic loop Text plus ElevenLabs speech
/voice1 agent_voice_active=True Keyboard ElevenLabs on assistant lines
/voice2 voice_mode_active=True, agent_voice_active=False Mic loop Text only from assistant
/voiceoff both False — Console confirmation

Text-to-speech (ElevenLabs)

  • speak_response streams TTS from ElevenLabs using VOICE_ID and ELEVENLABS_API_KEY.
  • speak_agent_response swaps to NYX_VOICE_ID or G1_VOICE_ID for agent-colored room lines.

Speech-to-text

  • speech_recognition.Recognizer with the default microphone.
  • recognize_google path (networked).

Boot-time voice latch

listen_for_voice_command(timeout=5) listens for the word voice to auto-start /voice mode; otherwise it prompts the operator to type /voice later.

Custom vocal shortcuts (MFCC)

custom_words maps spoken phrases to local audio files for similarity-triggered actions. Default paths are developer-specific (E:\Agents\Test 1\...); replace with your own media or remove unused entries.

Help bell

display_help() in help.py loads system_sounds/helpbell.mp3 relative to help.py.

Configuration checklist

Set in .env:

  • ELEVENLABS_API_KEY
  • VOICE_ID, NYX_VOICE_ID, G1_VOICE_ID
  • G1_VOICE_LIVE for /ask g1 live

Ensure the default Windows recording device matches expectations.