STT Callback

Speech-to-text Callback allows you to have your submitted audio processed asynchronously.

callback string

Deepgram’s Callback feature allows you to supply a callback URL to which transcriptions can be returned. When passed, Deepgram will immediately respond with a request_id before processing your audio asynchronously.

Enable Feature

To enable Callback, when you call Deepgram’s API, add a callback parameter in the query string and set it to the URL to which you would like transcriptions sent:

callback=URL

To transcribe audio from a file on your computer, run the following cURL command in a terminal or your favorite API client.

curl \
  --request POST \
  --header 'Authorization: Token YOUR_DEEPGRAM_API_KEY' \
  --header 'Content-Type: audio/wav' \
  --data-binary @youraudio.wav \
  --url 'https://api.deepgram.com/v1/listen?callback=URL'

URL Structure

An example URL is https://example.com/callback.

Your callback URLs may reference the following protocols:

  • For pre-recorded audio: http or https
  • For streaming audio, http, https, ws, or wss:

Authenticating Callback Requests

Authentication ensures the security and integrity of callback requests. There are two main methods for authenticating callback requests: using Basic Auth and utilizing the dg-token request header.

Using Basic Auth

You may embed username-password authentication credentials in the callback URL in the format https://username:[email protected]. However, it's important to note that only ports 80, 443, 8080, and 8443 are permitted for callbacks.

🚧

Only ports 80, 443, 8080, and 8443 are permitted for callbacks.

Using the dg-token Request Header

Alternatively, the callback request itself contains a header named dg-token. This header is automatically set to the API Key Identifier associated with the API Key used to submit the original request. This method provides a secure and straightforward means of authentication.

Results

Depending on the type of submitted audio, the response will vary.

Pre-recorded Audio

When Deepgram has finished analyzing the audio, it will send a POST request to the provided callback URL with an appropriate HTTP status code.

If the HTTP status code of the response to the callback POST request is unsuccessful (not 200-299), Deepgram will retry the callback up to 10 times with a 30 second delay between attempts.

If you would like Deepgram to make a PUT request rather than a POST request to your callback URL, you can add the callback_method=put query parameter in addition to callback=URL. Refer to the Using AWS S3 Presigned URLs with the Deepgram API guide for more information.

Streaming Audio

As Deepgram analyzes the audio, the way in which it sends requests back to the provided callback URL varies:

  • If your callback URL begins with http:// or https://, then Deepgram will send POST requests to the callback server for each streaming response.
  • If your callback URL begins with ws:// or wss://, then Deepgram will establish a WebSocket connection with the callback server and send WebSocket text messages that contain the streaming responses.

⚠️

If a WebSocket callback connection is disconnected at any point, the entire real-time transcription stream is killed; this maintains the strong guarantee of a one-to-one relationship between incoming real-time connections and outgoing WebSocket callback connections.


What’s Next