The pipeline

1. Get audio file(s) from the user
2. do representaion learning and clustering
3. use the made clusters to go back and train the neural network again
4. next switch the network to become a gan
5. train the gan at first to increase the amount of patterns in the data
6. next train the gan to make predicting the data hard while still making sure to keep the reconstuction loss low so that the patterns are captured
7. preprocessing, format the output in such a way so that the output format is matched