This repository contains the hackathon project for team Weird AI, the group that earned third place at Globant's Hack the Now & Next: The Future of Media.
Imagine you could automatically have the lyrics for a beautiful song written before your eyes. Using machine learning based in Python with TensorFlow, Weird A.I. built a working prototype of a system that is trained to generate lyrics in the style of Weird Al Yankovic. Tags can be used to direct the topic of a song, and the React-based front end is capable of fetching generated lyrics in quiz format to see if a user can tell a difference between "real" lyrics written by Weird Al and the "fake" lyrics crafted by Weird A.I.
A demo that uses previously arrays of previously generated lyrics can be found below. Note that the lyrics were entirely made with our backend, and a nearly complete API can be found wthin this repository.
https://weird-ai-c3e96.firebaseapp.com/
Note: we don't currently have the TensorFlow model backend running for live text generation. But the demo above has a list of songs created by our 1st generation model. Try typing in one of the following keywords to see some examples:
- AI Revolution
- Hackathon Blues
- Coffee
- Tech
- Eat
- Albatross
- Globant
- Kitchen
- Python
- TensorFlow
- C#
- React
- SCSS
- Redo web-scrapping to include helpful annotations such as verse/chorus labels and end of line markers
- Retrain model on new lyrics database
- Tune new model, but if results are unsatisfactory:
- Train a separate model for verses and one for choruses
- Implement algorithm to assemble songs line-by-line pulling from the 2 separate models as appropriate
Jonathan Leack, Front End Developer - https://github.com/JonnyBeoulve
Ivan Sued, Back End Developer - https://github.com/thefenry
Nathaniel Watkins, Data Scientist - https://github.com/TheNathanielWatkins
Chelsey Lew, Designer - https://github.com/Aliandrita