Tutorials

How to receive multilingual meeting transcription with the Recall.ai and Gladia integration

Amanda Zhu

October 19, 2023

We’re thrilled to introduce the partnership between Recall.ai and Gladia. Recall.ai is the unified API for meeting bots that work across Zoom, Google Meet, Microsoft Teams, and other video conferencing platforms. Gladia provides automatic speech-to-text recognition built on an optimized version of Whisper. Gladia has also extended upon Whisper with features such as real-time, multilingual transcription, and word-level timestamps.

This partnership allows developers and organizations to easily extract structured transcription data from meetings.

Integration Tutorial

In this tutorial, we will walk you through the step-by-step process of integrating Recall.ai with Gladia.

Step 1: Get API credentials

First, you’ll need to generate a Recall.ai API key in your account. If you don’t have a Recall.ai account, you can create an account here.

Step 2: Send a bot to your meeting

Recall.ai supports a number of video conferencing platforms, including Zoom, Google Meet, Microsoft Teams, Webex, Slack Huddles and more. For this tutorial, let’s use Zoom as an example.

Call the Recall.ai `Create Bot` API endpoint to send a bot to a meeting, with the `meeting_url` set to your Zoom link, and the `transcription_options` parameter set to `Gladia`. This will send a bot into the Zoom meeting that will pipe the audio to Gladia for transcription.
   
curl -X POST https://api.recall.ai/api/v1/bot \
    -H "Content-Type: application/json" \
    -H "Authorization: Token REPLACE-WITH-API-KEY" \
    -d '{"meeting_url": "REPLACE-WITH-MEETING-URL", \
         "transcription_options": {"provider": "gladia"}}'

After sending this request, you will receive a response with an `id`. Save this `id` for the next step!

Step 3: Get multilingual transcription

The final step is to get transcription from the meeting. Call the Recall.ai `Get Transcript` API endpoint with the following request.

curl https://api.recall.ai/api/v1/bot/REPLACE-WITH-ID/transcript \
    -H "Authorization: Token REPLACE-WITH-API-KEY"

In the response, we get back the transcription with speaker names, in a json format. This transcription output starts from the moment we sent the bot into the meeting, up until the point we sent the `Get Transcript` request.

[
  {
    "words": [
      {
        "text": "Hello",
        "start_timestamp": 1.55,
        "end_timestamp": 1.665
      },
      {
        "text": "Jean-Louis!",
        "start_timestamp": 1.87,
        "end_timestamp": 1.905
      }
    ],
    "speaker": "Amanda Zhu"
  },
  {
    "words": [
      {
        "text": "Bonjour",
        "start_timestamp": 2.03,
        "end_timestamp": 2.065
      },
      {
        "text": "Amanda!",
        "start_timestamp": 2.27,
        "end_timestamp": 2.305
      }				
    ],
    "speaker": "Jean-Louis Quéguiner"
  }
]

And that’s it! The Recall.ai and Gladia integration makes it ridiculously simple to get multilingual transcriptions from meetings. There is no need to specify language codes ahead of time.

Going beyond

Getting transcription is just the tip of the iceberg. After you receive the transcription, you may also want to summarize the content, extract action items, generate speaker talk times, and more - which are all things you can do using Recall.ai and Gladia together.

For more information on what else you can do with this meeting data, feel free to check out the comprehensive API documentation provided by Recall.ai and Gladia.

Sign up for Recall.ai here, and sign up for Gladia for free here.