@dave: thanks for watching! Are you referring to something like a machine following a real musician? That sounds really cool! I’ve made a note of it to try it out sometime.

I’m trying to pull out the database related work into another repo: https://github.com/ssrihari/kosha

At this stage, it’s mostly just scraping a few websites on the internet and stitching the data together. But it already extensive data around compositions, ragas and their metadata, and has some fair querying capabilities.

The next step that I think would prove to be difficult, is extracting only vocal/main-instrument frequencies from audio. A lot of the recordings have poor quality, so I’ll have to programmatically detect that and modify the frequency extraction appropriately.

If you’re interested in contributing to either of these repos (synthesis or database), you’re more than welcome! We can catch up over hangouts for context. At the moment though, I’m trying to put together some solid documentation around what the intentions are, and what the basic design so far is.