I have several legally obtained recordings ... too many. They're not labeled and I would like to see if I can "scan" these recordings for a particular voice so I know if that person is on that particular digital recording.

All legal, as I am party to all conversations concerning legal matters that I am suing the other parties for.

I wonder if there is a free program that will recognize say "bills" (fictitious name) or "henrys" (again fictitious), or whomever so I can find bill or henry or whomever in specific recording, or even my own voice or my wife's.

I have a legal battle ahead for me that the legal aid foundation is going to take because of the merit of my case, but I still have to collate the evidence.

Any ideas?

3 Answers 3


The most relevant piece of software I could find is called Praat. It's free, and it offers spectral, pitch, and formant analysis—but it was developed for the study of phonetics.

This is deep software though, and, though I feel comfortable in max, reaktor, and other visual programming environments, Praat is beyond me. It's the most relevant software i've been able to find for your needs, but it's not for the meek!

good luck!


3 options:

  1. Izotope RX4 in demo mode, is capable of detecting 'similar sounds'. I haven't used in on voices, but it could work. But you need to consider that it needs clean samples (in the same recording) and that demo will only run for 1 month…

  2. Dragon Dictate is a dictation app that allows discrete dictation in iOS and Mac OS X (maybe also other OS'es). It works very well on clean sources. You let it listen to the source and it 'writes everything' down. You would have to search on specific terms to get your info.

  3. The human ear is capable of listening to speed up sounds to about 4× the original speed. You could simply listen to the recordings and detect the voices yourself. It will take less time than normally and will be more certain that you get complete results. Using RX4 or another program could possibly skip important parts of the recording.

… or you can use all of the above :)


However you do it, you'll probably need a clean sample of the sounds(voices/words) that you want to target, as a spectral reference.

The sample could be taken from the recording itself, as long as it was a clean sample(no other sounds involved). This way the sonic characteristics of the recording device would be included. For that reason, the sample would be recording specific, especially if they are noisy recordings or in varying environments.

Just something to think about.

Not the answer you're looking for? Browse other questions tagged or ask your own question.