Understanding AI Transcription in tawk.to

The instructions below are for desktops and laptops only.

AI Transcription lets AI agents receive, transcribe, and respond to voice messages automatically. This feature is part of the AI Assist add-on and works across all channels that support audio attachments or voice messages.

How it works

AI Transcription works on any channel that supports audio attachments or voice messages:
  1. The customer sends a voice message.
  2. The AI agent converts it to text and generates a response.
  3. The response is sent back to the customer.

Supported audio formats and limits

AI Transcription supports the following audio formats:
  • MP3
  • MP4
  • MPEG
  • MPGA
  • M4A
  • WAV
  • WEBM

The following limits apply:
  • Maximum file size: 25 MB per audio file
  • Maximum duration: 5 minutes per voice message

These limits help maintain reliable performance and transcription accuracy.

Voice transcription and message credits

Voice transcription uses AI Assist message credits, just like other AI Assist features, such as AI Commands and Smart Reply. You are charged for transcription based on the length of the spoken audio, not the full file length. Silent sections are automatically skipped, and partial minutes round up to the next full minute (for example, 61 seconds counts as 2 minutes).

Credits are calculated as follows:
  • Transcription: 1 message credit per started minute of speech, capped at a maximum of 5 credits per voice message
  • AI reply: 1 message credit per AI-generated response

If the same audio file is already transcribed and stored for the same AI agent, reusing the file does not use additional transcription credits.

You can track your transcription credit usage in the Reporting section of the dashboard, where you can see each transcription and view historical usage.

Examples

A 30-second voice message uses:
  • 1 credit for transcription
  • 1 credit for the AI-generated reply
Total: 2 message credits

A 1 minute 35 second voice message uses:
  • 2 credits for transcription
  • 1 credit for the AI reply
Total: 3 message credits

A 4 minute 20 second voice message uses:
  • 5 credits for transcription (cap)
  • 1 credit for the AI reply
Total: 6 message credits

To learn more about managing credits and overages, see this guide:
How to top up AI Assist message credits

Supported languages

AI Transcription supports 99+ languages, including:
  • English
  • Spanish
  • French
  • German
  • Italian
  • Portuguese
  • Chinese
  • Japanese
  • Korean
  • Arabic
  • Hindi

The spoken language in each voice message is automatically detected. No manual language selection is required.

How to enable or disable AI Transcription

1. Log in to your tawk.to account.

2. Select the correct property.

3. Click Add-ons on the left navigation bar.

4. Click Settings under AI Assist.

5. Select your AI agent.

6. In the Settings tab, scroll to the bottom of the page, and turn the AI Transcription toggle on or off.

Changes apply immediately after you turn the switch on or off. If the AI agent is assigned to a channel that supports voice messages (for example, a WhatsApp inbox) in the Channels section of the Settings page, the agent will be able to respond to voice messages as soon as AI Transcription is enabled.

Additional considerations

  • AI Transcription only applies to conversations assigned to an AI agent.


  • Disabling AI Transcription stops voice-to-text processing but does not affect text-based messages.

  • AI Transcription uses the same message credits pool as other AI Assist features, such as AI Commands and Smart Reply.

Related guides


If you have feedback about this article, or if you need more help:

Was this article helpful?

1 out of 1 liked this article

Still need help? Message Us