How to Transcribe an Audio File with Python and AssemblyAI

How to transcribe an audio file with Python and AssemblyAI
Share on social icon.Share on social icon.Share on social icon.Share on social icon.
This introduction paragraph is special. Why is it special? Well, if you're 10
moves ahead of me like a chess Grandmaster, then you've probably already guessed
it. This introduction was not hand Typed. It was spoken and then transcribed
with Python using the Assembly AI API.

Seriously. Don't believe me? Below is the audio recording of my sweet, sultry voice creating that paragraph. I spoke it off the cuff, and it only took me 2 takes to not stumble over my own words that I imagined ūüėÖ.

I would have spoken this entire post by voice, but because of my limited human memory, I would have to script it in writing, and that defeats the proof of concept, now doesn't it? Here's how you can transcribe your audio files with Python too.

Sign up for an AssemblyAI API Token

You can sign up for a free AssemblyAI account in seconds by just entering your email address. After you verify your account from your email address, you're taken right back to your new account where you can see your API token in your dashboard.

Find an audio file to try it on

AssemblyAI transcription supports loads of file types, and gives you 5 hours of audio transcription per month absolutely free, with no billing setup or commitment. So find that perfect mic check audio file you want to try it out on, you'll have plenty of chances to test it with our API.

Upload your file for transcription

If you already have a publicly accessible URL for your audio file already hosted somewhere, you can skip this step... but if not, the first step is to upload your audio file to AssemblyAI with this code:

import sys
import time
import requests

filename = '/path/to/your/audio_file.mp3'

def read_file(filename, chunk_size=5242880):
  with open(filename, 'rb') as _file:
    while True:
      data =
      if not data:
      yield data

headers = {'authorization': 'YOUR_OWN_API_KEY'}

response =
    headers=headers, data=read_file(filename)


A successful upload will yield a JSON response similar this:

{"upload_url": ""}

Submit the audio file for transcription

The code below uses the upload_url key value from the JSON response in the previous code, so it will run sequentially with that. If you already have a url for your audio file, replace it with response.json()['upload_url'].

endpoint = ''

json = {
    'audio_url': response.json()['upload_url']

heads = {
    'authorization': 'YOUR_OWN_API_KEY',
    'content-type': 'application/json'

resp =, json=json, headers=heads)

A successful response will look like so, displaying the "status" : "queued"‚Äć

    # keep track of the id for later
    "id": "5551722-f677-48a6-9287-39c0aafd9ac1",
    # note that the status is currently "queued"
    "status": "queued",    
    "acoustic_model": "assemblyai_default",
    "audio_duration": null,
    "audio_url": "",
    "confidence": null,
    "dual_channel": null,
    "format_text": true,
    "language_model": "assemblyai_default",
    "punctuate": true,
    "text": null,
    "utterances": null,
    "webhook_status_code": null,
    "webhook_url": null,
    "words": null

Get that transcription you've been longing for!!

This code below will check the status of the transcription, and concatenates the transcription id resp.json()['id'] into the url for you, so it will once again run sequentially with the previous code.

status_point = '' + resp.json()['id']

status_header = {'authorization':'YOUR_OWN_API_KEY'} 

status_check = requests.get(status_point, headers=status_header)


You'll see in the above code we are printing 3 things. The first print statement print(status_check.json()['status']) prints out the status of the transcription, which goes from "queued" to "processing" to "completed". Audio files take around 25% of their audio duration to complete, so a 10 minute audio file would complete within 2.5 minutes. You'll want to run this above code in a loop until the status key shows "completed".

Once the status is "completed", the second print statement print(status_check.json()['text']) will print out the actual transcription text. And the third print statement prints out the entire API response, with a bunch more meta data like the timings for when each word was spoken, the confidence for each word, and a bunch of other data! You can see a complete example of the API's response JSON on the AssemblyAI Docs here.

My mom did manual transcription work in the 90's, and she hated it. REJOICE at this beautiful API that AssemblyAI has created for you, and all the headaches it can save you.

The full sequential code can be found in this colab notebook here! Happy transcribing!

Subscribe to our blog!

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

You may also like

Checkout some of our recent research and product updates

Python Speech Recognition in Under 25 Lines of Code
Python Speech Recognition in Under 25 Lines of Code

How to build a YouTube Downloader in Python
How to build a YouTube downloader in Python

How to get the transcript of a YouTube video
How to get the transcript of a YouTube video

In this blog post, I'm going to show you how to build a command line tool that will download a video from a YouTube link and extract the transcription for you via AssemblyAI in Python 3!


Unlock your media with our advanced features like PII Redaction,
Keyword Boosts, Automatic Transcript Highlights, and more