How to use voice input

Speak a prompt into the AI panel with browser recognition or local Whisper processing.

Voice input requires a network connection when the Whisper model loads. Voice processing remains local after the model loads for that session.

1. Open Preferences, enter the API key required for AI submission, and choose browser recognition, local Whisper, or off. Start voice input is unavailable until a key is present and voice mode is not off.
2. Open AI Conversation and select Start voice input. The browser requests microphone access and the control changes to Stop voice input while listening.
3. Stop input and review the recognized text before submitting it. The prompt remains editable and no diagram change occurs until submission is accepted.

Recognition uses one English locale, and the local Whisper model is English-only in 0.2.5.

Result

Recognized text is ready for review in the AI prompt field.

Related

• How to generate with AI
• Understanding privacy and storage
• Browser support