Automatic Vowel Analysis script for Praat

GitHub repository

Download the script

This script is designed to do acoustic analysis of vowels on a whole linguistic corpus.

How to use it

  1. Open Praat
  2. Click the “Praat” button in the upper menu
  3. Open Praat Script
  4. Select the script
  5. Click the “Run” button in the upper menu
  6. Follow the instructions provided by the script

NOTE: PRAAT TENDS TO FLICKER A LOT WITH THIS KIND OF SCRIPTS, BE CAREFUL IF YOU ARE PHOTOSENSITIVE I don’t think I can do much about it, this really is just a Praat thing. If this is a problem for your workflow, feel free to write me an email. There is technically a way to avoid the flickering, its just a bit of a pain in the ass and I cannot really do it for you in the code, I have to teach you how to bypass the buggy interface from the terminal.

Additional instructions

You should have 2 folders somewhere on your PC.

The first one should contain the corpus, that is, .wav and .TextGrid files with corresponding names. The second one should be ideally empty and ready to be filled with the results generated by this script. Or you may just do it all in a single folder…

For me it’s a bit too chaotic, but if that is what you want, just input the same path twice. The script will not stand in your way.

Example of use

You may try it with the example audio + TextGrid pair attached in the GitHub repository. For a bit of context, it is a young Spanish woman reading in an anechoic chamber. Her dialect employs 5 vowels /a/, /e/, /i/, /o/ and /u/. The segmentation was done automatically using MFA, so don’t expect human level precision, however, it is perfectly serviceable for this example. If you want to simulate the script analysing a whole corpus, just copy and paste the pair a few times in the same folder