Are there programs that take a recording of human speech as input, perform some kind of phonetic analysis and then return a graphic readout of the phonemes that have been found?

I'm currently trying to familiarize myself with basic articulatory phonetics and I'd been hoping to find some means of objectively verifying whether I can reproduce the sounds correctly.

Preferably this program would be Open Source and able to be run on Linux.

1 Answer 1


You are talking about creating a spectrogram. One program that does this is Praat:


Another is WaveSurfer:


Praat is widely used, so you will find a lot of helpful documentation on it. There are also a lot of free Praat plug-ins and scripts available for download.

WaveSurfer allows you to see the spectrogram (and waveform, and other visual feedback) for an input sound in "real time", that is, as soon as you start speaking into the microphone you start seeing the spectrogram scrolling across the window (see this related question/answer). It also allows you to create spectrograms in color (another related question/answer).

  • 1
    I actually think they're asking for a printout of phonemes, rather than a spectrogram. Digital, not analog. Pretty much a hopeless dream at the moment -- one would hafta translate into English spelling and thence into phonemic representation if one wanted to use offtheshelf software,
    – jlawler
    Sep 7, 2014 at 17:05
  • 1
    @jlawler haha! I abstracted away from some of the terminology used in the question, but after rereading it I think you're probably right. I'll wait to hear from the OP and remove my answer accordingly if that is indeed the case. Sep 7, 2014 at 19:22
  • You're right, that's indeed what I had in mind. It seemed unlikely to me that this kind of software would exist and work well, but this had been mentioned by someone in a video, so I wanted to ask in spite of my incredulousness. Apart from that, since I'd like to get a basic grounding in phonetics, familiarizing myself with Praat might be well worth the effort anyway. Sep 11, 2014 at 11:28
  • @user1428640 Not only does what you describe not exist, but such a hypothetical system would not be useful for the purpose you outline. It might guess correctly the phoneme you are trying to produce but that wouldn't mean you are producing it "correctly" for the language and context in question. Better to find out from an expert what acoustic features are crucial and to learn how to spot those features in a spectrogram. Sep 13, 2014 at 3:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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