A full stack AI app that generates (musico_nn, backend), streams (frontend), and labels (frontend) music.

It is certain that artificial inteligence (neural networks) is developing fast in these years. However, the application of it in people's daily life is far behind the accademic progess. Musico suggests a business model that may be possible for any generative neural network. Musico sells the outputs of neural networks and collects data from users. The backend is a recurrent neural network that generates and stores music everyday in the Google cloud. The front end is a Android app that streams music for the user. It also collects "likes" and "dislikes" of music from the user, which are labels that can be used for improing our neural network gradually. In other owrds, the good outputs picked by users can be fed into neural network to train further. It's free for listenning. This app could make profit by selling the copyright of the hit music.

Backend provides API to let other AI developers sell music generated by their neural network.

a. Library/Framework

TensorFlow, Flask, Google Cloud Platform, Android Studio, ExoPlayer

b. Neural Network

LSTM

a. Type

abc notation - the text-based music notation system

b. Source

796 songs from http://abcnotation.com/

a. Backend

Music collection https://musico.depplenny.com/

Genrating music manully (takes 1 minute) https://musico.depplenny.com/musician

b. Frontend

Android App https://musico.depplenny.com/static/musico.apk

c. Train RNN

musico_nn/