Chat
Talk to any model.
Voice in, voice out. Claude, GPT, Gemini, whichever you pin.
A voice for every mood.
Pick the voice you want to hear. Keep the AI model you've been working with: they're separate choices.
OpenAI voices
Claude voice
Gemini voice
Google voices+ open-source voicesFeels like a real conversation.
Start talking and it stops talking, the way you'd interrupt a friend who's rambling.
It picks up on the natural pause at the end of your sentence. No buttons to hold.
Walk, drive, cook: the conversation keeps going. The transcript waits on your screen.
What you get
Speech-to-text on your turn, natural TTS on the reply. No typing required.
Dozens of TTS voices across providers. Pin one per agent, change per chat.
Talk over the assistant to cut it off: like a real conversation.
Voice mode works with whichever LLM you pinned. Switch models without leaving voice.
Every voice turn is saved as text in your chat. Search later, share, branch.
Who it's for
Think out loud on the walk in; the transcript and reply wait on your screen.
Dictate a draft, have it read back, revise without touching the keyboard.
Hands-free, eyes-free interaction with every model idapt supports.
How it works
- 1Tap the mic
Voice mode opens full-screen with a quiet listening state.
- 2Speak naturally
Pauses end your turn. The transcript appears as you talk.
- 3Listen back
The reply streams in voice. Tap the screen to cut it off.
- 4Continue or close
Take the conversation to text any time. History keeps both sides.
Frequently asked
Take a walk. Keep thinking.
The transcript will be there when you sit back down.