3 ms·
Generating “happy accidents” are my primary interest when it comes to using ML/AI for music composition. 8 times out of 10 I can generate my own but I’d love to
by maroonblazer 8y ago
Generating “happy accidents” are my primary interest when it comes to using ML/AI for music composition. 8 times out of 10 I can generate my own but I’d love to be able to cover that last 20%. This seems like a sweet spot for current technology.
Any chance you’ll open source your approach?
- tabtab 8y agoI find diminishing returns. If I try to add logic and gizmos to get it to automatically be closer to my ideal, it just complicates things and would probably take 10x longer to "finish". Thus, a "perfect" composer may be do-able, but the end result would be a Swiss-Army Spruce Goose. That is if the sheer complexity doesn't introduce bugs to make it worse. As far as open-sourcing it, I'd have to clean it up to be presentable to humans-that-are-not-me, being it's a ball of ad-hoc experiments.