Friday, July 04, 2025

Manipulating bird song

We've probably all heard the novelty Christmas songs with dogs barking or cats meowing the notes; and if we're old enough probably remember Marvin Suggs and his Muppaphone.

A bark can approximate a note in the scale, and if it isn't quite there you can tune it to fit, and with enough samples you could program your synthesizer to bark out whatever tune you cared to. All you need to worry about is the note, duration, and volume, and a good synthesizer will know how to deal with key velocity. You have an alphabet; you can plug in whatever sounds you want.

But what about birdsong? What would you need to take into account if you wanted to compose music using recorded bird chirps, cheeps, chips, warbles, etc?

Bird chirps run faster than human-sung notes. In the time it takes me to sing "la" the bird runs through a trill and a chirp to top it off, even shifting back and forth between notes. Maybe it crescendos, rises and falls. The warbling won't be the same frequency from one species to another.

Matching one "note" to the next isn't like picking letters from an alphabet, but matching textures, speeds, even which sets of notes were used in what isn't a chord. It's more like picking characters from a Chinese typewriter.

You could get around that by slowing the bird calls way down, trimming out a clean part, and "autotuning." It would probably sound interesting, and maybe even good, but I doubt it would sound much like the birdsong it came from.

Maybe the Chinese typewriter is the way to go. To get started, suppose we limit ourselves to 2 octaves of the primary note: 24 primary notes. No warble or warble at a set frequency (we can fiddle with the sound to fit one of our set values): say 8. The secondary note: probably a choice of 8, but some will sound nice together with different ones. Rise and fall shape: 5. Hmm. So far that's 7680 different sounds in the sound bank, with 4 descriptors. I hate to think how many calls you'd have to listen through to settle on that number.

Of course you can take different parts of a single call to represent different ... I can't call it a note, because it's more than just the note ... complexes. Still, even without more descriptors, that's a lot of time somebody (you) has to invest to sort out the complexes, before you start playing.

OK, start simple. Sort out a single type of sound you want to work with (warble type, flat "rise and fall", secondary note difference from the primary). That'll be maybe several weeks of work to get 24 notes (getting a feel for the territory takes a while), and then you program your synthesizer and try it out.

I'd bet that it wouldn't sound quite clean right out of the gate--that attack will be hard to make smooth; and probably some things I didn't think of will jump out at the musician who tries to play it for the first time.

Maybe start even simpler: take some bird chirps, split them into sections, and find the primary frequency for each--if there is one. If there isn't, trying to shoehorn them into notes may not be the way to go. You might have fun trying to assemble your own bird calls, though.

Would it be good enough to be an instrument in its own right, or merely a novelty?

No clue.

No comments: