2

Suppose you have multiple waveform snippets (mostly 1-2 seconds long each), and they all "somehow" fit together (i.e., are the respective beginning/end of each) => complement each other.

However, it is not clear in what order.

Is there some class of software tools that can "match" the wave forms that fit best to each other and reassemble the snippets to make the most coherent sense ?

Metaphor: Waveforms as "puzzle" pieces. Software to piece the most fitting ones together, repeat 'til a most-complete puzzle is finished (need not be 100%, just 'as good as possible')

TL;DR: Image stitching but then for music snippets.


As (expertly) put by Bob:

However, I suppose you could have software that will analyse the smoothest ordering of such snippets, where the volumes and frequencies of adjacent beginnings and endings matched most closely - this would be quite interesting to look into.

THIS. is pretty much what I'm after, a class of software that does exactly this. :)

6
  • 1
    Do the waveforms overlap? What are you trying to do? Where the snippets you have come from? Commented May 3, 2021 at 17:01
  • yes, some overlap. These are multiple instances of one and the same song (with silence between the snippets), but the snippets differ in length and position in relation to the track. (Imagine "punching out" parts of a song in regular intervals, but in two versions, so you get kind of a 'complementing jigsaw'. With bith versions creating the whole song. See example in OP post.
    – Rye
    Commented May 3, 2021 at 17:15
  • I don’t think you’ll be able to find software that will be able to tell you “musically” which is the best order for these snippets, as this is largely a matter of taste, with no “correct” solution. However, I suppose you could have software that will analyse the smoothest ordering of such snippets, where the volumes and frequencies of adjacent beginnings and endings matched most closely - this would be quite interesting to look into. Commented May 3, 2021 at 17:27
  • Well, but technically, the waveforms - as long as they are similar enough (if not even the same) could be "assigned" with fuzz. I'm not even talking about "perfect" match, just "close enough" EDIT: YES! "matched most closesly", that is pretty much what I'm asking about! [Remember how you can put music into "spectrograms" for analysis ? I'd recon one could perhaps figure out a way to "match" with that as basis. ]
    – Rye
    Commented May 3, 2021 at 17:30
  • 4
    What is your actual goal here? To stitch snippets of a song you cannot source in its entirety back into one coherent piece? or…?
    – Tetsujin
    Commented May 3, 2021 at 17:36

0

Browse other questions tagged or ask your own question.