Voice Mode: Real-Time Voice Conversations
Talk to AI in real time Voice Mode is a full screen, real time voice conversation with AI. Speak naturally and get spoken answers back instantly, like a phone call. It is not just transcription: th...
Talk to AI in real time
Voice Mode is a full-screen, real-time voice conversation with AI. Speak naturally and get spoken answers back instantly, like a phone call. It is not just transcription: the assistant can also use tools mid-call, from web search to generating images, video, and music, or acting on your connected apps.
How to start
- Open any chat
- Make sure the text input is empty (the waveform button only shows when nothing is typed)
- Click the audio waveform icon in the chat input
- Allow microphone access when your browser asks
- Voice Mode opens full-screen. Start talking
Voices
Pick from 6 voices. Ash, Echo, and Verse are deeper; Alloy, Coral, and Shimmer are brighter.
- Tap the play button next to a voice to preview it before the session starts
- Ash is the default. Your choice is remembered for next time
- Switch voices between sessions whenever you like
What you can do in a call
| Feature | Details |
|---|---|
| Real-time conversation | Speak normally and interrupt mid-sentence; the AI keeps context for the whole call |
| Photo context | Tap the camera button to share a photo and ask questions about it |
| Photo editing | Ask it to restyle or modify the photo you shared |
| Media generation | Ask for an image, video, or song; it appears on screen and is saved to the chat and Keep when the call ends |
| Web search | Ask about current events and it looks them up live |
| Connected apps | Use your app connections by voice once they are linked in Additions |
| Voice badge | The session is saved to your chat history marked “Voice” |
You can even name a model out loud (“use Kling”, “make it with Sora”) when asking for media. Anything generated inside Voice Mode is billed at that tool’s standard rate, on top of the per-second voice rate.
Cost and billing
- 0.25 credits per second, so 15 credits per minute
- The first 5 seconds (connection setup) are not billed
- Maximum 30 minutes per session (about 450 credits). The call closes at the cap; start a fresh one to keep talking
- A live counter shows elapsed time and credits while you talk
- Voice Mode holds an upfront reservation (up to 500 credits, or your whole balance if lower) and refunds what you do not use when the call ends. Hanging up instantly costs nothing
- You need at least 5 credits to start. On a low balance, the call simply ends when the reserved credits run out
Requirements
- A paid plan: Pro, Ultra, or Max, or a seat on a team whose owner has one
- Microphone permission in your browser
- A stable internet connection
Use cases
- Brainstorming: talk through ideas and get instant feedback
- Language practice: a patient conversation partner
- Hands-free work: AI help while your hands are busy
- Quick questions: faster than typing
- Visual questions: share a photo and discuss it
Tips
- Speak clearly at a natural pace; you can interrupt it mid-sentence
- Attach a photo for anything visual (“What is this?” or “Remove the background”)
- Watch the live counter to keep spend predictable
- Try different voices between sessions to find the one that fits
Related: Understanding credits, text-to-speech, the voice changer, and plans and pricing.