MAI Realtime 1
Speech-to-speech model with natural conversational flow and precise steerability
MAI Realtime is unavailable right now. Please try again later.
MAI Labs is a limited preview of new and experimental AI models. AI can make mistakes.
Start a live voice conversation using your microphone. Sign-in is required.
Finish signing in to continue. Connecting Ready, start talking Listening Speaking Muted Ending in %ds Mute Unmute Microphone muted. Microphone unmuted. Session ended. This demo allows sessions of up to 2 minutes. Conversation closed. Cancelled. MAI Realtime is unavailable right now. Please try again later. The selected Playground hasn't enabled a MAI Realtime 1 connection. The service configuration needs to be updated before you can start. MAI Realtime isn't available for this account. Sign-in didn't finish. Allow pop-ups for this site and try again. Your sign-in expired. Sign in again to continue. Accept the Terms of Use to continue. The Terms of Use couldn't be checked. Try again. Your consent couldn't be saved. Try again. Microphone access isn't available in this browser. Allow microphone access in your browser to start a conversation. No microphone was found. Check your microphone and try again. Audio playback couldn't start. Try again. Sorry, our model is busy. Please try again later. Connection timed out. Try again. Can't connect right now. Check your connection and try again. The conversation disconnected unexpectedly. Start again to reconnect. Couldn't finish that response. Please try asking something else. This conversation cannot continue. Start a new conversation. The conversation couldn't be started. Try again. MAI Realtime 1 uses your microphone for a live voice conversation. Before you start: StartListens, understands context and responds with human-like fluency.
Accuracy on spoken reasoning questions from the Big Bench Audio benchmark.
Per minute, with reasoning costs charged separately. At Input it's $0.01, and at output it's $5 per 1M tokens.
P50 Latency on BigBench Audio
Performance
Dig deeper
Register your interest
Our speech-to-speech model is in preview. Tell us what you’re building, and we’ll find the right next step for your team.