Every run out of music or wanting something new? Now you can make your own tunes, no experience required.

What it does

Procedurally generate a song sequence using a neural network trained with a variety of songs. Analyze .midi files and encode into a string file that can then be understood by our program. Convert the string file back to both midi and mp3 for playback on other devices.

How I built it

A lot of trial and error after successfully implementing the network to ensure that the network functioned as expected.

Challenges I ran into

Tuning the neural network to work well with our sample data.

Accomplishments that I'm proud of

It doesn't sound that bad :)

What I learned

Implementing a neural network, as well as performing various sound processing techniques on music.

What's next for

Crowd-sourced learning.

Share this project: