Need help with How_to_generate_music_in_tensorflow_LIVE?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

134 Stars 74 Forks Apache License 2.0 5 Commits 7 Opened issues

Services available


Need anything else?

Contributors list

# 13,411
5 commits



This is the code for this video on Youtube by Siraj Raval as part of the Deep Learning Nanodegree course with Udacity. We're going to use an encoder-decoder model to generate a MIDI sequence given an input sequence of MIDI ragtime songs.


The program requires the following dependencies (easy to install using pip): * Python 3 * TensorFlow (tested with v0.10.0rc0. Won't work with previous versions) * CUDA (for using gpu, see TensorFlow installation page for more details) * Numpy (should be installed with TensorFlow) * Mido (midi library) * Tqdm (for the nice progression bars) * OpenCv (Sorry, there is no simple way to install it with python 3. It's primarily used as visualisation tool to print the piano roll so is quite optional. All OpenCv calls are contained inside the imgconnector file so if you want to use test the program without OpenCv, you can try removing the functions inside the file)


To train the model, simply run
. Once trained, you can generate the results with --test --sample_length 500
. For more help and options, use
python -h

To visualize the computational graph and the cost with TensorBoard, run

tensorboard --logdir save/


Credits for the code go to Conchylicultor. I've merely created a wrapper to get people started.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.