MixList

CSE 481I: Sound Design capstone project for Josh Bean, Jeremy Cruz and Gerard Gaimari

Installing Audacity

Before everything can work together you need to download the source code for Audacity and install it with mod-script-pipe enabled. There are instructions for downloading the source code and installing it here. There is also some information about enabling mod-script-pipe here. In order to get it to run on Windows we had to build and run from Visual Studio. For some reason running Audacity from the executable file wouldn't allow mod-script-pipe to work even though it was enabled.

Also since we are using Spotify you need to add a config.py file to the mixlist/analyzer/analyzer folder before pip install .. This config file should contain two variables:

CLIENT_ID = <your client id>
CLIENT_SECRET = <your client secret>

Top Level

• main.py runs the GUI for MixList. By default it will create a cache in the folder you are running it from called mixlist_cache. There it will store all the serialized analyses for files that it has analyzed.
• pipe_test.py is called from main.py to test the connection to Audacity. For some reason, after mixing one song and pressing Test Audacity it doesn't work correctly although you can still mix.

Folders/Modules

• analyzer contains the code for the analyzer portion of MixList and also defines some of the data structures used throughout the pipeline
  • NOTE: Before running you'll need to cd analyzer/ and run pip install ., the analyzer uses a lot of third party libraries so it's best that we put those dependencies in setup.py. There were a number of issues configuring the eyed3 module however we reverted the version to be installed to not import the python-magic library which was causing issues.
• composer contains the code for the MixList composer.
• composer_scripts was a folder containing some of the test scripts we used when developing the composer.
• keyfinder contains code for a keyfinding algorithm. It's a fairly simple module since it can just use the LibROSA command for getting the constant q transform of an audio file which otherwise would be a fair amount of code. We didn't end up integrating this into our platform because the results weren't great.
• optimizer contains the code for MixList optimization.
• proposal contains the writeup of our proposal and a script to compile it with pandoc
• scripts contains some scripts that we used during the development process