Saving Transcripts From Your Terminal

I've recently started doing a lot more work directly in my terminal - and I've learned that writing Bash scripts doesn't have to be scary. Today, we'll write a set of commands and scripts to execute directly in our terminal.

Before You Start

You will need a Deepgram API Key - get one here. You will also need to install jq.

Making a cURL Request

Open your terminal and type (or copy and paste) the following, not forgetting to change YOUR_DEEPGRAM_API_KEY with a real API Key, and then press enter:

Let's break down each part of this request:

• --request POST: is a HTTP request with the POST method.

• --header 'Authorization: Token YOUR_DEEPGRAM_API_KEY' - include details to link this request with our account/project.

• --header 'Content-Type: application/json' - JSON data will be sent with this request.

• --data '{"url":"https://static.deepgram.com/examples/nasa-spacewalk-interview.wav"}'. - is the JSON data sent to Deepgram (an object containing one url parameter).

• --url 'https://api.deepgram.com/v1/listen?punctuate=true' - the URL to make the request to (Deepgram's endpoint). punctuate=true enables the punctuation feature.

• \ allows us to break one command over several lines for readability.

Shortening Your Request

The first example is handy for understanding all of the required parameters. Here is a more concise way to make the same request:

The first thing you'll notice is that the URL comes immediately after the curl command. You may have also noted the absence of an HTTP method - this would normally default to a GET request, but as this request has a body, a POST request is inferred. --header is shortened to -H, and --data to -d.

Adding jq

jq is an excellent command-line utility that allows you to display and manipulate JSON data. On the terminal, a pipe (|) is often used to send the output of one command as an input for a second command. jq expects some JSON as input and an expression to describe how to display it.

This jq expression will extract just the transcript from the returned data object:

| jq '.results.channels[0].alternatives[0].transcript'

You can add it to the end of your cURL request like so:

Saving Output to File

Once you have the correct data extracted and formatted from jq, you can redirect the output to a new file by appending > output.txt to any command that prints to the terminal. Here it is in practice:

Processing Multiple Files

You can create .sh files to execute from your terminal, which contain multiple lines of bash script. Create a new file called transcripts.sh and open it in a code editor. Copy and paste the following:

Let's break down each part of this file:

• The first line - #!/bin/bash - is a shebang, and specifies which program should be called to run the script. In this case, bash.

• urls is a variable containing an array with three URLs. Notice that arrays use parentheses, and items are separated by a space only.

• dg_features and dg_key are variables you should alter for your exact use case.

• Inside of the for loop:

  • filename extracts the last part of the URL (the filename), which will later be used to name the output file.

  • The curl command is the same as before, with variables interpolated. The output is stored in a new variable called RESPONSE.

  • RESPONSE is sent to jq, and then redirected into a new text file.

Run the file in your terminal with ./transcripts.sh. As a note, this request uses Deepgram's punctuation, utterances, diarize, and tier features.

Playing with jq

jq is a remarkably powerful tool. The following expression will loop through the results.utterances array, and format a string for each item interpolating the speaker identifier and transcript text:

The output looks like this:

I hope you found this valuable and interesting. If you have any questions, please feel free to get in touch - we love to help!

If you have any feedback about this post, or anything else around Deepgram, we'd love to hear from you. Please let us know in our GitHub discussions .