Skip to content

AdaptiveStep/watson-beat

 
 

Repository files navigation

The Watson Beat

Using machine learning to spur human creativity is so 2018! Watson has used a combination of reinforcement learning and neural networks to create the Watson Beat.

Inputs

  • a simple 10-20 second melody file in midi format
  • parameters for the creativity engine

Outputs

  • Several layers of music in midi format that Watson created using your input as inspiriation

Getting Started

This repo, is all the code for the The Watson Beat. It all runs locally using the terminal. To run it, there are a few things you need set up on your local computer first, like python 🐍!, and a basic familiarity with git.

Set you machine up for dev - if you are a first timer!

Here are some directions to get your machine set up to run code, or just google it! You will also need to copy this code repository to your computer,

Use Git

To get the code on your computer you will need to git clone this repo. Here is some great documentation on git for new comers. Also, if you want to make contributions to the code base, we are very willing to review merge requests (see forking). All pull requests and commits to branches intended for release must acknowledge the Developer Certificate of Origin by including the one liner: DCO 1.1 Signed-off-by: {Full Name} <{email address}> (without the {})

Install Project Level Dependencies

So, you just spent half an hour setting your machine up for python, and I'm telling you there is more set-up? Yep! Welcome to the wonderful world of code. Open your terminal, and navigate to the directory you created with your git clone command. If you do an ls you should see this file, README.md, and the specfic python requirements for this project are in requirements.txt.
You can install the specific python packages you need for your Watson Beat Client with the following command (again, make sure you are in this directory):

pip install -r requirements.txt

Ready to Run

There are two main ingredients to prepare to create your song with The Watson Beat. First, you need a midi file with a simple melody. Best results come when you keep this short, about 10 seconds. Also, leave some space between your notes to give the creativity engine some wiggle room.

The second ingredient is the ini file. The ini file give the creativity engine all of the spices to use, including time signature, "mood," and tempo. Read this to learn more about ini files.

Usage:
usage: wbDev.py [-h] [-i INIFILE] [-m MIDIFILEPATH] [-o OUTPUTPATH] [-u]

optional arguments:

  -h               show this help message and exit

  -i INIFILE       Store the ini File

  -m MIDIFILEPATH  Midi File Path

  -o OUTPUTPATH    Path were all the output mid files are stored (default ./output/)

  -u               usage

To run, clone this git project in your filesystem if you haven't already. Open a terminal window, navigate to the project home $WB_HOME (for example, this might be /Users/Moe/watson-beat)

cd $WB_HOME/src

Now, in the current working directory, there is a python script that will call the code, wbDev.py with the parameters you pass it. To get help, just pass it the -h flag:

python wbDev.py -h

If you pass it no paramaters,

python wbDev.py

it will use the default Ini file $WB_HOME/src/Ini/Space.ini, the default midi file $WB_HOME/src/Midi/mary.mid, and output all the files to ./output/

To pass in specific midi and ini files use the following syntax:

python wbDev.py -i Ini/ReggaePop.ini -m Midi/mary.mid -o /Users/Moe/midifiles/

Putting it together

This is where you get to have some fun! Now, you have a set of midi files, you can use your favorite audio tools to apply virtual instruments to the sections and mix them together into a song.

HAPPY COMPOSING!

Video Tutorials

How to Change the Length of a Composition

Importing MIDI Files

How to Customize Moods Part 1

How to Customize Moods Part 2

License

See LICENSE.txt

Original Authors: Janani Mukundan, Jeremy Hodge, and Richard Daskas
Assign Pull Requests to: amchaney

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%