Exploring Async Deepgram API: Speech-to-Text using Python

Today will explore the Deepgram API for converting voice to text [transcription]. Whether building a voice assistant, transcribing meetings or creating a voice-controlled app, Deepgram makes it easier than ever to get started.

What is Deepgram?

Deepgram is a powerful speech recognition platform that uses advanced machine learning models to transcribe audio in real-time. It offers an easy-to-use API that developers can integrate into their applications for tasks like transcribing phone calls, converting meetings into text, or even analyzing customer interactions.

Why Use Deepgram?

• Accuracy: Deepgram boasts high accuracy rates thanks to its deep learning algorithms trained on vast datasets.

• Real-Time Transcription: Get instant results as you speak, perfect for live applications.

• Multiple Languages: Supports several languages and accents, making it versatile for global applications.

Getting Started with Deepgram API

Install - pip install httpx

Importing Required Libraries

import httpx
import asyncio
import logging
import traceback

Defining the Asynchronous Function

#recording_url: The URL of the audio file to be transcribed.
#callback_url: The URL to which Deepgram will send the #transcription results (optional).
#api_key: Your Deepgram API key.

async def transcribe_audio(recording_url: str, callback_url: str, api_key: str):
    url = "https://api.deepgram.com/v1/listen"

    # Define headers
    headers = {
        "Authorization": f"Token {api_key}"
    }

    # Define query parameters
    query_params = {
        "callback_method": "post",
        "callback": callback_url
    }

    # Define body parameters
    body_params = {
        "url": recording_url
    }

4. Sending the Asynchronous Request

    logger.info(f"Sending request to {url} with headers: {headers}, query: {query_params}, body: {body_params}")

    async with httpx.AsyncClient(timeout=60.0) as client:
        try:
            # Make a POST request with query parameters and body
            response = await client.post(url, headers=headers, params=query_params, json=body_params)
            response.raise_for_status()  # Raise an error for HTTP error responses
            result = response.json()
            logger.info(f"Response received: {result}")

            return result

We create an instance of httpx.AsyncClient with a timeout of 60 seconds. Using async with ensures that the client is properly closed after the block is executed.
If the request is successful, we parse the JSON response and log it, then return the result.

Call back URL :

You can use for sample call back URL for testing.

conclusion:

This structured approach highlights how to utilize asynchronous programming in Python to interact with the Deepgram API efficiently. By breaking the code into blocks and explaining each part, readers can better understand the implementation and how to adapt it to their own needs.