concat_chunks (praat script)
concat_chunks is a Praat script that automatically concatenates audio chunks from a wav + textgrid. It can be useful to illustrate the principles of speech synthesis by concatenation. Students can easily create a rudimentary system with their own voice in Praat. The concatenation is done by merely copy-pasting the audio chunks one after the other, therefore the output is obviously of low quality. It can only be useful for illustration purposes, as it gives students a quick hands-on experience of how speech synthesis works.
The concat_chunks Praat script can be downloaded at the bottom of this page, as well as some sample files and a powerpoint introduction to synthesis by concatenation.
What you need
In order to run concat_chunks, you need to download Praat and the concat_chunks.praat script. Additionally, you will need 3 files.
Create a recording in .wav format with one or more sentences containing all the phonemes for your language. It may be useful to check out what a phonetic pangram is. Also check the sample provided for British English in recording.wav below.
Once you have a recording, create a textgrid in Praat and segment the audio into phonemes (you can use any set of symbols of your choice). Check the sample provided in recording.TextGrid below.
Finally, you need a .txt file in which you specify the phonetic transcription of the word or sentence that you wish to synthesize. The transcription should contain one phoneme per line. Check the sample provided in text.txt below.
For ease of use, I suggest you save your .wav file, .textgrid file, .txt file and concat_chunks.praat in the same folder.
NB: the .wav and the .textgrid files must have the same name.
How to run
Open Praat, then click the menu 'Praat', 'Open Praat script...' and select concat_chunks.praat. Click 'Run', then 'Run' again, and you will see this window:
If you followed the instructions above and are using the sample files provided, you can leave all defaults. Otherwise, simply insert the name of your .wav and .txt files (do NOT include extensions, as per the screenshot above). Click 'OK'; if everything goes as it should, you will find your synthesized sound in Praat's object list with the name 'chain' (for the given sample, the synthesized sentence is 'she tries to concentrate').