This module uses Google's Gemini 3.8 Live Extended Thinking model, the deep-reasoning variant of Gemini 3.8 Live built for complex voice tasks that need multi-step reasoning: press Start Conversation, allow microphone access, and just speak - the AI reasons and talks at the same time, and for questions that need checking or several steps it will first say something like "let me check that", then walk you through its progress and conclusion while the status line shows that the AI is thinking. You can choose the thinking level (low, medium or high) in the panel: a higher level reasons more deeply but takes longer to answer; medium is a good default for everyday conversation, and if you only need fast replies use Live Voice Conversation gemini-3.8-live instead. Barge-in, transcripts for both sides, background-noise filtering and automatic language switching are all supported as well. Great for spoken planning and decision analysis, tutoring and problem solving, step-by-step reasoning through hard questions, and any scenario where you want the AI to explain as it thinks. On a phone, the conversation stops automatically when the browser moves to the background, the screen locks, or the browser is closed. Headphones are recommended so the AI voice from your speakers is not picked up by the microphone. Audio input only - text input and file upload are not supported. Please use a recent Chrome or Edge browser with microphone support. Output content formats: Text/Audio Pricing: Medium-high This module supports continuous conversations. AI can make mistakes. Please verify important information.