Music Generative Model
This is a project I’ve been thinking about for a while and now I decided to start working on it. The aim is to generate piano music using a generative model. The reason why I decided to limit the scope of the project to piano music is that I play piano and I understand piano more than other instruments.
In addition, piano is a wide pitch range and polyphonic instrument so it can imitate most of the instruments. Except this…
Anyway, there is a lot of potential of piano music generation. One of the possible use case in the future is to generate piano music cover of a song even there are a lot of instruments and vocal involved. For example, Animenz is a YouTube channel that uploads his piano music cover of anime songs. His arrangement is insanely good and I think it will be cool if the model can reach this level(or even half).
The reason why I started this blog series is to document the progress of my research and share my thoughts with others. So to avoid the situation where I can’t finish the project, I will post the outcome on the Internet by the end of October, whether it is good or bad.
In the next blog post I will dive into some of technical details of the project. Stay tuned!
Updates from Future
Seems like I gave up documenting this project due to many events in my life. Everything is fine, just that it is harder for me to squeeze out time for it. I did work on this project but I found it very hard to do it on my local environment (limited storage space and 4GB VRAM). I believe that I will still continue this project in the future, and I am looking forward to use cloud GPU for experimenting and testing by that time
There is a discussion thread in EleutherAI discord group about symbolic music generation. Check it out if anyone is interested. asfjkl