Recognising/removing music from MP3 tracks

Discussion in 'Digital Audio' started by Dreamkatcha, Mar 5, 2006.

  1. Dreamkatcha macrumors regular

    Dreamkatcha

    Joined:
    Mar 17, 2003
    Location:
    Manchester, England
    #1
    Hi,

    I was wondering if it's possible to look at the wave form of an MP3 track and decipher which parts are music with lyrics and which parts are speech. You see, I record a lot of talk radio and would like to be able to cut out all the music. It's not that I don't like music, just that I can listen to it whenever I want so don't want it interrupting debate shows.

    Maybe there's a piece of software out there that can do this automatically?
     
  2. thumper macrumors 6502

    thumper

    Joined:
    Oct 25, 2003
    Location:
    Under the Sea
    #2
    huh?

    are you asking how to seperate the vocals from the music? cuz if you are
    you cannt really do this... you can try EQing or phase.. but the end product will sound like crap.

    or are you asking how to remove the commercial breaks,which have music.
    if thats the case.. then its easy to see where music starts and end since the radio compresses
    it like mad. the wave form of the music will look like a brick wall.
     
  3. zimv20 macrumors 601

    zimv20

    Joined:
    Jul 18, 2002
    Location:
    toronto
    #3
    that strikes me as a manual process. but as thumper says, it'll only work if they're separate sounds in the timeline.

    if you zoom out to the right point, you can get used to a particular track pretty quickly. "oh, there's the verse. and there's the bridge." shouldn't be too hard, if i'm understanding what you want to do.
     
  4. CanadaRAM macrumors G5

    CanadaRAM

    Joined:
    Oct 11, 2004
    Location:
    On the Left Coast - Victoria BC Canada
    #4
    You're trying to cut chunks of time out of the track, right, like skipping commercials on a videotape?
    As posted, I think this'll be manual, never heard of software that can automatically distinguish content that is music from content that is voice. However, you (with your superior carbon-based processing unit that excels at fuzzy pattern matching) should be able to see easily by looking at the waveforms where the breaks and music are.
     
  5. Dreamkatcha thread starter macrumors regular

    Dreamkatcha

    Joined:
    Mar 17, 2003
    Location:
    Manchester, England
    #5
    Yes, I'm trying to chop out complete segments like you would when skipping commercials - some of the broadcasts I listen to are a combination of music and talk and since I listen to music all day long at work I'd rather not hear the same stuff in my DIY podcasts. If I leave the music alone and then try to fast forward through it while I'm walking about I often skip to the next track by accident and lose my place. Very irritating.

    When looking at the wave forms I did notice some solid chunks of sound peaks, but then looking at the other channel threw me a bit because the patterns were dissimilar (clearly I'm no audio expert). Through sheer guess work I imagined there would be some kind of synchrony between the two. I'm sure if, as you suggested, I zoom in I'll be able to see things more clearly.

    I'll give it a go and see if all the manual editing is worth the trouble - I listen to shows and then trash them straightaway so it's not as though there's a long-term benefit.

    Thanks for your help everyone. :)
     
  6. zimv20 macrumors 601

    zimv20

    Joined:
    Jul 18, 2002
    Location:
    toronto
    #6
    "the other channel"? do you mean you're looking at a stereo pair or two mono tracks?
     
  7. beatsme macrumors 65816

    beatsme

    Joined:
    Oct 6, 2005
    #7
    well...

    I'd load the track in GarageBand and view the waveform there. There's usually some dead air before and after commercial breaks, so you should see the signal flatline at those spots. Cut and paste...
     
  8. Dreamkatcha thread starter macrumors regular

    Dreamkatcha

    Joined:
    Mar 17, 2003
    Location:
    Manchester, England
    #8
    The MP3s I'm talking about are recorded in stereo and I'm using Audacity to edit them (or WavePad in Windows when I'm at work - great free audio editor btw).

    Dead air? *gasp* That's a major radio faux pas isn't it? :D Good tip though, thanks.
     
  9. WinterMute Moderator emeritus

    WinterMute

    Joined:
    Jan 19, 2003
    Location:
    London, England
    #9
    Problem is that all commercial radio is heavily compressed (limited actually) and that takes out any variations in level in the waveform, as the DJ's voice is actually at the same level as the music.

    Unlimited broadcasts would be very difficult to listen to in noisy environments, which tends to defeat the purpose of commercial radio!

    I think the only way to do this would involve something like Beat Detective being able to recognise the rhythmic component, not the loudness.

    I'm sure some clever boy could come up with something in Max/MSP to do the job.

    (not me though...)
     
  10. zimv20 macrumors 601

    zimv20

    Joined:
    Jul 18, 2002
    Location:
    toronto
    #10
    i'd imagine an application of FFT could make the distinction, assuming the music has a wider frequency range than human speech.
     

Share This Page