Voice input

You can dictate a message. A separate speech recognition model turns the recording into text. Regular chat models don't accept audio, so it has to be set up separately.

Setup

1. Connect a service that has speech recognition models: OpenRouter, OpenAI, Groq or Mistral AI.
2. Expand the Voice input step in Settings → Overview (or open Models → Dictation) and choose the service and model.
3. Turn on Dictate messages.

Until recognition is set up, the input field shows a Set up voice input button instead of the microphone.

How to dictate

1. Click Dictate a message (the microphone icon) and speak. One recording can be up to 15 minutes.
2. Click Stop. MindDuck recognizes the recording.
3. The text appears in the input field. Check it and send it.

If you turn on Auto-send, the message goes out right after recognition, without checking.

Cancel discards the recording without sending anything.

What is stored where

• The audio is sent only to the chosen recognition service.
• The recording is saved as an attachment of the message and is available on other devices. You can play it right in the chat. A message dictated on another device is marked Message dictated on another device.

Messages