You can dictate assistant messages with the microphone button in the composer. Speak, tap Done, and your words appear in the message box so you can edit or send as usual.
How to use voice input
Open the AI assistant from the sidebar or the Agent page.
Tap the microphone button next to the send control.
Allow microphone access when your browser asks.
Speak clearly. A live waveform shows while you are recording.
Tap Done when you are finished. Your transcript is added to the composer.
Review the text, then send the message like any other prompt.
Voice dictation uses AI credits the same way as typing a message and sending it.
Private chat
If Private is on in the composer, recorded clips are transcribed with the same private processing as your other assistant messages.
Browser support
Chrome, Edge, and Safari support live speech recognition when your connection allows it.
Brave and other browsers that block live speech still work: the clip is recorded locally and transcribed securely after you tap Done.
Firefox does not support voice input yet. Use Chrome, Edge, or Safari, or type your message.
Tips
Keep clips short (under about a minute) for the best results.
If nothing was captured, try again and speak a little louder.
You can edit the transcript before sending.
Common questions
Why is the mic disabled? Voice input needs Chrome, Edge, or Safari. Firefox is not supported yet.
Why do I need to tap Done? Done stops recording and adds your words to the composer so you can review them before sending.
Does dictation send the message automatically? No. You still tap send when you are ready.