audio-speed

This is a program that measures the speed of speech and notifies whether too fast or too slow.

Using Assembly AI async endpoints, this pplication will measure the speakers speaking rate and notify them if they talk too quickly or slowly. Greater than 160 words per minute is too fast, and less than 120 is too slow.

This application ia a terminal application. Therefore, in order to run this application, this repository has to be cloned to your local machine.

Requirements

• ubuntu/redhat
• python3

---

Clone repo

$ git clone [email protected]:dnjoe96/audio-speed.git

Enter into the repo

$ cd audio-speed

Run the setup script script.sh, providing the API auth key from Assembly AI

audio-speed$ ./script.sh <auth-key>

Now your environment is setup with the right dependencies and the authentication key is setup in the system environment variable.

• Note: This script has to be run if the system is restarted.

---

Now you can run the script

audio-speed$ ./app.py -f <path to audio file>

example : -

audio-speed$ ./app.py -f audiofolder/media.mp3

---

or

audio-speed$ ./app.py -f <url of audio file>

example: -

audio-speed$ ./app.py -f https://facebook.com/some-audo-file

---

Here is an test with an audio file locally

(venv) somename@somehost:~/audio-speed$ ./app.py -f audiofolder/media.mp3 
file uploaded
url posted
queued
processing
processing
processing
processing
processing
done
speed is = 135.48
Speed ok
(venv) somename@somehost:~/Desktop/Project25/audio-speed$ 

---

Implementation

I use the GET POST method on the Assembly AI API

endpoint = "https://api.assemblyai.com/v2/transcript/"

with the POST method, we pass in a json payload containing the audio url like

{
  'audio_url': 'https://some-url.com'
}

to the endpoint and get a json object as response containing keys. At this stage, my key of interest is the id

Because with the id, I make a GET request to the same endpoint

endpoint = "https://api.assemblyai.com/v2/transcript/{}".format(_id)

Getting a response of error or completed means that the request is done processing and we can analyse the json object to get the return data.

The speed is gotten from

speed =  length-of-words / audio-duration

For more information on the modules and API used, here are some resources.

• Assembly AI docs
• Click python module

---

AUTHOR: Joseph Emmanuel || Linkedin