Gemini Live Api Webrtc Integration

Lets have a natural conversation with your AI inside your pocket.

2025-09-16 17:16:23.953 INFO (MainThread) [custom_components.gemini_live_hass_integration.llm.gemini] [Wakeword 'ok_nabu'] detected with score 0.861
2025-09-16 17:16:35.474 INFO (MainThread) [custom_components.gemini_live_hass_integration.llm.gemini] FUNCTION CALL: id='function-call-696465183208074604' args={'name': 'light'} name='HassTurnOn'
2025-09-16 17:16:35.474 INFO (MainThread) [homeassistant.helpers.intent] Triggering intent handler <OnOffIntentHandler - HassTurnOn>
2025-09-16 17:16:35.480 INFO (MainThread) [custom_components.gemini_live_hass_integration.llm.gemini] Function Response: {'speech': {}, 'response_type': 'action_done', 'data': {'targets': [], 'success': [{'name': 'light', 'type': <IntentResponseTargetType.ENTITY: 'entity'>, 'id': 'input_boolean.light'}], 'failed': []}}

Why Choosing Gemini Live Api Plus Webrtc?

  • Fast Response Time < 0.5s : Almost identical to talking or response to people in real
  • Integrated VAD: Voice Activity Detection on gemini socket fast and accurate
  • Interrupt detection: The response stream can be interrupted through VAD
  • Free to use : The live model can has at most 3 concurrent session lasting forever
  • Echo Cancellation: The speaker and microphone will not interfere
  • Automatic Gain Control (AGC): Adjust the volume and output (from client side)
  • Noise Suppression: Reduce background noise
  • Tool Use: Both synchronous and asynchronous function calling is available
  • [NEW] Open Wake Word: Can wake up anytime or switch into active standby mode
  • OpenAI
  • Gemini Live Api
0 voters
2 Likes