mimikit-notebooks

Colab Badge GitHub Badge

The mimikit-notebooks are a small collections of jupyter notebooks containing scripts to train audio generative neural networks.

Notebooks Overview

Model

Plain

Colab

Wavenet

GitHub Wavenet

Colab Wavenet

SampleRNN

GitHub SampleRNN

Colab SampleRNN

FreqNet

GitHub FreqNet

Colab FreqNet

Seq2SeqLSTM

GitHub S2S

Colab S2S

You’ll find examples of what those notebooks output when run on different single audio files at

Since all notebooks follow the same design, this documentation covers the general usage of the notebooks. It is intended for people who have little prior knowledge of python & deep-learning and wants to get a hands-on first contact.

The fastest track, is to

  • open one of the “Open in Colab” above

  • evaluate the cells with shift+enter.

  • The program will require you to connect to your Gdrive. Before doing so, make sure you placed some audio files in <username>/MyDrive/data on your Gdrive.

  • Keep running the cells with shift+enter (Do not pay attention to the warnings and errors you get from the !pip install... cell, those are not critic).

  • Generated audio will be displayed during training.

If you have questions or troubles running the scripts, join our slack channel Slack Badge