I'm recording audio for a scientific experiment. Recording to MP3 with a smartphone or dictaphone is 'easy' enough (within audio quality constraints).

But is there any program which can convert MP3 audio files to XY (time,amplitude) data. Data would be available to saved in columns in ascii text e.g. CSV format.

If not I guess I'll just use an adafruit Mic+Preamp board & use an arduino as a datalogger, I just wanted to investigate if the MP3 route first.

  • 1
    Are you able to program? This should be pretty easy with Python.
    – Eric S
    Mar 9, 2018 at 14:45
  • yes. There are built-in methods to read mp3 in matlab & labview, but I wanted to check there isn't any thing already out there before going down the coding road. Mar 9, 2018 at 19:18

1 Answer 1


If you search using the terms "soundwave art," you'll find a number of programs, both online and offline that will generate images from audio sources.

The result is what could be considered a line graph, in any number of standard image formats.

I opened Audacity, a free audio editing program and loaded Hal9000 saying "I'm sorry, Dave" then chopped it down to "Sorry."

Sorry image

The above image is that one word, converted from stereo to mono, then expanded to fill the screen.

I found an online resource to take the screen capture and convert it to graphical data.

sorry data image

Because I had no valid reference for the data, I used arbitrary points to calibrate the image, resulting in values of less than one for the positive portion of the graph and greater than minus one for the negative portion of the graph. The numbers appeared in scientific notation, implying extremely small detail capability.

Using the View Data option provides for a glance at the raw numbers as well as the option to export to a file:

sorry data

If the level of manual labor doesn't scare you off, you can achieve your objective.

  • ooh, hadn't seen the webplotdigitzer before, I was still using the old 'xyextract' from some Uni which was always a bit clunky. your site looks much friendlier thanks. I hadn't thought about this graphical approach, handy for small chunks but yes potentially laborious for large datasets. Mar 9, 2018 at 19:21

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.