The Living Thing / Notebooks :

Concatenative synthesis

Transferring timbre from one sound to another; Synthesis by example. There are a lot of ways this could be done, but the classic is the “talking orchestra” vocoder, which always seemed ham fisted to me. I think about it more in terms of sprase_coding, but there are many ways to think about it. When you refer to “concatenative synthesis” or an “Audio mosaic”, you usually mean using a granular synthesis method. This being the epoch of neural networks, someone will probably get style transfer for audio functioning soon.

The most comprehensive overview of classic concatenativge style stuff IMO is contained in Graham Coleman’s doctoral dissertation, below, which frames it in terms of loss functions and descriptors. (Cole15)

There are a few classic implementations about;

Audio analogies

Related: analysis-resynthesis, learning gamelan.

Refs