Recently I have been experimenting with algorithmic music generation. As of now, its nothing fancy. Just a set of sequentially arranged algorithmically generated melodies in the foreground and a set of fixed background music pattern. Here is an example (just the music, the video is something else). But I don't really know how I can modify this to fit Indian ragas. Any souls here who has explored the relationship between swaras and frequencies? Everything is generated from Python. Example code: https://github.com/vinaynk/VLearnX/blob/main/python/projects/wavsynth/basicmelody_examples.py

Checkout glicol.org, tidal cycles, strudel, supercollider

Interesting. But audio sequencing is an easy problem even in Python. Languages like this and chuck do not bring much benefit for regular programmers. The hard part is generating melodies and harmonies algorithmically.

I played around with music generation 2 years ago using deep learning (RNN and LSTMs). It came out moderately well. Here's the repo https://github.com/Akilesh1989/music_generation_with_deep_learning and here's the output. https://soundcloud.com/akilesh-raj/jingle-bells-machine-generated?utm_source=clipboard&utm_medium=text&utm_campaign=social_sharing For the first few minutes, you'd wonder if this was Jingle bells but then it makes sense.

That sounds very natural. Any way to limit the repetition? I want to try deep learning too. Probably after trying simple algorithms. There is quite a lot to explore before trying data driven approaches.

Yes, you can. You just have to train accordingly, I think I only trained on 1 or 2 minutes of data.