Skip to main content
When your deployment has a transcription backend configured, a microphone button appears in the composer next to the send button. Recordings are transcribed server-side (Whisper-style speech-to-text) and the text lands in the message box — nothing is sent until you press send.

​
Dictate a message

1

Start recording

Click the microphone button (Start recording). The first time, your browser asks for microphone permission — allow it.
2

Speak

While recording, the button shows a pulsing stop symbol. Speak normally; you can dictate several sentences.
3

Stop

Click the button again (Stop recording). The button shows a spinner while the audio is Transcribing.
4

Review and send

The transcribed text is appended to whatever is already in the message box. Read it, fix anything the transcription got wrong, and send. Dictation never sends automatically.
Transcription uses your interface language as a hint, which makes short recordings in languages other than English noticeably more accurate. Set your language in the user menu.

​
If the microphone doesn’t work

If recording can’t start, IMP shows a Microphone unavailable message explaining why:
“Microphone permission was denied. Click the lock/permissions icon next to the URL and allow microphone access for this site.” — re-enable the permission in your browser’s site settings, then try again.
“No microphone was found on this device.” — plug in or enable a microphone, then reload the page.
“The microphone is in use by another application.” — close the other app (video call software is the usual culprit) and try again.
Recording requires HTTPS (or localhost). If your deployment is served over plain HTTP, browsers block microphone access entirely — ask your administrator.
If a recording fails to transcribe, IMP shows “Transcription failed” — your recording is discarded, so dictate again.
Dictation transcribes short voice input into the message box. To transcribe whole meetings or audio files, use Transcripts in the sidebar instead.