4

My parents have written short stories at a time where PCs were not so present. Unfortunately, many of those hand-written manuscripts seem to be lost. However, there are a lot of old cassettes which might contain the stories spoken.

We're about of getting rid of cassette players and we also want to dispose off all those old cassettes, but I'd like to preserve the stories if possible.

The labelling is far from perfect, so I'd probably need to go through all ~90 of them, record them into my PC with Audacity and then find the stories. Of course there's also a lot of music recorded.

Therefore I need a software that could help me to distinguish spoken text from music in audio files (MP3). Ideally it would generate a picture or similar which classifies the sounds. And it should have an integrated player so that I can directly jump to relevant positions.

Example classification: red = music, green = text, grey = silence:

Example output

I'd need that with a ready to use UI for Windows and it should be gratis.

I found the Phython script smacpy on Github, but that requires training the software. As far as I understand it may only classify whole files and not parts inside a file. pyAudioAnalysis (Github) might classify audio parts, but does not have a UI for easy usage.

The solution does not need to be fully automated, but should give clear hints. From what I saw in Audacity it's hard for me to tell text and music apart. But if there's a plugin for Audacity, that would be ok, too.

  • 4
    I would suggest first capture & backup the audio files - that is the one thing that needs done ASAP because once they are gone they can never be recovered. Then why not try smacpy - it is free - and first train it with half a dozen files that you have hand classified then try it with the rest - if you have to leave it running for a day or two so be it. – Steve Barnes Aug 22 '16 at 6:18
  • @SteveBarnes: I have already started recording. Maybe you're right. I wouldn't totally trust the software, so I would proof-listen some of them anyway. Perhaps I should give it a try. – Thomas Weller Aug 22 '16 at 19:07
1

If you use Audacity to view various aspects of the recordings, it is likely that the difference between music and spoken word will be very visually apparent.

  • 1
    I guess the asker wants an automated solution, though. Examining each track on Audacity probably takes as much time as listening on fast-forward. – Nicolas Raoul Aug 22 '16 at 2:54
  • 1
    Based on his phrasing ("Ideally it would generate a picture or similar which classifies the sounds. And it should have an integrated player so that I can directly jump to relevant positions") it doesn't seem to me he needs a fully automated solution. – spruceb Aug 22 '16 at 4:46
  • 1
    I have updated the question. I use Audacity for a long time already. Maybe you just need to tell me a) the settings to use and b) how to identify the difference. It doesn't need to be a fully automated solution. I think it is possible to open 90 files in Audacity manually, if the rest is straight forward. – Thomas Weller Aug 22 '16 at 19:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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