No Code Transcription: Simplifying Workflows with Deepgram and Make.com

Recently I was using my mobile phone to message back-and-forth with someone in a language that I speak well but not fluently. The person preferred to send me voice messages, using the microphone to record himself speaking and then sending me the audio file.

Normally this would be a fine way to communicate, but as someone who doesn’t speak the language fluently, I found myself wishing I could read his messages along with having the option to listen to them.

This of course gave me an idea for how I could use Deepgram to make my life easier.

Deepgram has just released an integration with Make, giving me the perfect tool to build a no-code automation solution to turn those audio messages into text messages.

In this blog post, I’ll walk you through how I created the automation. I chose Telegram as the messaging app, but you could do this with WhatsApp or another messenger app of your choice. Once you’ve built one tool like this in Make, you’ll be ready to create all sorts of automations.

Check out the Make + Deepgram documentation to learn more about getting started with Deepgram’s Make integration

Here’s the scenario (i.e. workflow) we’ll build in Make:

The workflow will consist of:

1. Step 1: Telegram Watch Updates - Triggers the workflow to start by watching a specific Telegram account to see any changes, such as messages being sent to that account.
2. Step 2: Telegram Download a File - Downloads the message in Telegram so we have the data of that message to work with (we’ll want the audio data from a voice message).
3. Step 3: CloudConvert Convert a File - Takes the raw audio data from the telegram message and converts it into a form that can be used by Deepgram.
4. Step 4: Deepgram Transcribe a Prerecorded Audio File from URL - Takes the audio file and transcribes it into text.
5. Step 5: Telegram Send a Text Message or a Reply - The final action of the workflow which sends the text transcription as a message to the Telegram account.

I’ll walk you through each step so you can learn how to set up the integration for this specific scenario.

For the three integrations you’ll be using in Make - Telegram, CloudConvert, and Deepgram, you’ll need an API Key. Here’s how to get set up with each.

You will need to create a Telegram bot so we can test the workflow.  Telegram has extensive documentation about working with Telegram bots, but you can also just follow these steps.

Create your Telegram bot by messaging @BotFather on Telegram. Send the message “/start” to instruct BotFather to walk you through creating your first bot.

You’ll need to create a new bot with the command /newbot. Then you can get your API token with the command /token.

Be sure to message your new bot so that you have the conversation readily available to work with as you test the Make workflow.

First sign up or login at cloudconvert.com. Once you are logged in, click on your profile to head to the dashboard. From there you can navigate to “Authorization” in the left side navigation menu. Create a new API key and be sure to save it.

Next, sign up or log in at deepgram.com.

If this is your first time signing up for Deepgram you'll get $200 in Free Credit (up to 45,000 minutes), absolutely free. No credit card needed!

Navigate to API Keys from the left side navigation menu and create a new API key.

Last but not least, sign up for your account at Make, and from the dashboard, click on “Create a new Scenario”. Now you’re ready to get started with building the workflow!

In this section, you'll learn how to build a Make scenario that combines several different integrations, or "modules" to produce one automated task.

You’ll start by choosing a trigger. The trigger is the first event in the workflow because it “triggers” the automation to move through each step of the workflow when some initial event happens.

Click on the big plus sign and search for Telegram Bot. Then select “Watch Updates”. This will prompt you to create a webhook. Name the webhook whatever you want, then click on “Create a connection” to add your API token. Save it and click “OK”.

Test Step One

Test this step by clicking “Run once.” This will start the workflow so that it is now watching your Telegram account with the Telegram bot. Send a message to your Telegram bot from another Telegram account (I just used my personal Telegram account to message the bot). In Make, you should see that the workflow sees the message and gives you an output now that something has occurred with the linked account.

You can click on the bubble above the module to see information that came through. You should be able to see the message that you sent! (It can be found under Message > Chat > Text)

The next module will also use Telegram, but this time you’ll search for the “Download a File” action.

Connect with your API Token just like you did in the previous step. You will also be prompted to add a File ID. You’ll need to take this from the output of the previous step. Click into the input and notice the popup that shows all the output options. Click into Message > Attachment > File ID to select the File ID.

Test Step Two

In order to test this step, click on “Run once”, which will run the entire workflow. Just like in step one, you can send a message to your telegram bot to trigger the workflow. However, this time you must send a voice message for the test to work! That’s because the File ID that the module is waiting for only comes with a voice message. If you send a text message, the test will fail.

Click on the plus sign to add CloudConvert as the next module. Search for “Convert a File,” and then create the connection by adding the CloudConvert API key you created earlier (it should be a very long key).

In the form, select “upload a file” as the Input File. For “Convert Options”, choose “I don’t know the input type” and then choose an output format. You’ll need to choose a format that is accepted by Deepgram, such as wav or mp3.

For file name, you can choose whatever you want, but be sure to add a file extension in the name.

Then in “Conversion and engine specific options”, add a name (whatever you want) and a value (must be the Data output). The Data output is the voice message that came from Telegram in the previous step.

Test Step Three

Click “Run once” and then send a voice message again to your Telegram bot to trigger the workflow. You will see CloudConvert working to convert the file. The output should look something like this:

You’ll use the Temporary URL in the next step when you set up the Deepgram module.

Now we can set up the Deepgram module. Click on the plus sign and search for Deepgram among the list of integrations. Choose the action “Transcribe a Prerecorded Audio File from URL.”

Connect to Deepgram by adding your API Key:

Then fill out the form. You will need to add the Temporary URL from the previous CloudConvert step in the URL input. Then select any Deepgram configuration options you would like to add. I recommend selecting “Nova” for the model and changing Smart Formatting to “true” so that the transcript is formatted with punctuation and paragraphs.

Test Step Four

Click “Run once” and then send a voice message again to your Telegram bot to trigger the workflow. You will see Deepgram working to create the transcription. When the output becomes available, click into the “Results” section to find the transcript:

You’re almost to the end! The last thing to do is add another Telegram module that will send the transcript as a text message. Search for the Telegram action “Send a Text Message or a Reply.”

Fill out the form so that the Chat ID contains the “Message: Chat: ID” from the first step, the “Watch Updates” module. This will ensure that the text message is sent to the original Telegram account that was used at the start of the workflow.

Then add the Deepgram Transcript in the “Text” input. Be sure to drill down into Results > Channels > Alternatives to find the Transcript.

Test Step Five

Now you can run the entire workflow to see the result. Click “Run once” and watch as each step completes. If everything is successful, not only will you see the success message for each module, but you’ll also receive the text message in your Telegram bot!

In my test, I used my personal Telegram account to send a message. Make ran the workflow when the message arrived and then converted it to a text message for me. That text message was delivered by the Deepgram Telegram bot!

This step-by-step guide has walked you through the process of using Deepgram and Make to automate the creation of speech-to-text transcriptions without having to write any code. Now that you are familiar with this tool, I recommend that you explore all the integrations available and start putting your own ideas into action. 

And if you come up with some interesting ways to use Deepgram, please share them in our Discord or GitHub Discussions communities!

We encourage you to explore Deepgram by checking out the following resources:

1. Deepgram API Playground 
2. Deepgram Documentation
3. Deepgram Starter Apps