Pixels-n-Pitches

Uses machine learning to take in audio files, analyze their style, and display them as pixel art.

Comment

Inspiration

Data Visualization, Love of Music, Love of Visual Art,

What it does

This takes songs in the form of midi files and uses AI to make pixel art based on those songs

How we built it

We created images with songs and scales to give the AI information for training. We used Python to build the AI and train it.

Challenges we ran into

We ran into a few challenges with sizing the images properly to have the pixel output not be too large or small so that our program was manageable to run. Our program is also limited by the amount of data that can be put in because some midi files are longer than others and a lot of our calculations are based on the numbers or measures or notes as input data.

Accomplishments that we're proud of

We had 2 people in the group who knew nothing about AI going into this so we are proud to have something that works at the end of this hackathon even if it isn't as polished as we originally wanted it.

What we learned

We learned that even if you have little to no knowledge in a technology, if you apply yourself you can learn a lot in just a short amount of time, especially when working together. We learned a lot about machine learning, and how to manipulate Midi files.

What's next for Pixels-n-Pitches

It would be fun to experiment with different data with visualization tools. Finding different ways to visualize data in ways beyond just pixel art. We would also like to change the output so that it can recognize pitches too so the images change based on the octive the songs are in. We also want to try and figure out why most of the images that are output have pixels that form vertical bands across the image.

Built With

Share this project:

Updates