Hallo,
does anyone else have music that by far doesn’t sound as nice as the initial example?
Not just different, but just by far not as jazzy, or clear.
I’m wondering what I do incorrectly.
That’s a common observation. One issue is the quality of the MIDI player that is included in the assignment. Its voices are really bad.
Certainly not a hit, but a really cool experience to generate something that with more training may become music. The reason your (and mine too) generated music was not as we could have expected is because in the lab we are playing with very little training data and probably very small models. There’s enough data and enough of a model to generate something, but if you want to generate something more decent, you’ll need lots of data, lots of training, and probably a much more robust model. But now you know where to get started!
Juan