Have a song in your heart and a tune on the tip of your tongue... ! Make a fun app to enhance the creativity, play around and generate music using Machine Learning.
What it does
- Generate short & fun music clips based on your key selection.
- Currently genre selection is - Classic, Pop, Hip-Hop & Electro
- Feeding the network musical data in ABC notation, it generates a sequence character-by-character, whose output is a musical composition in ABC notation. This output is fed into an ABC-to-MIDI converter to produce an audible music file.
How I built it
Pytorch + FastAI + Android Neural network in PyTorch using LSTM layers.
Accomplishments that I'm proud of
- Generated music through sequential generation of characters from a recurrent neural network.
- Network consisted of LSTM layers for learning information about song structure, musical measures.
- Deploy a pytorch model on android
What's next for Musiclii
- Currently it is free to use for a 20sec generated music.
- Improve the ML model, for users to select custom time length.
- Further customize generated music with recorded user audio input like "This is fun", "Cool" etc.