Data-Driven DJ

by Brian Foo

FAQs

I like lists; I will answer most of these questions with a list.

What are your goals for this project?

  1. Make good music - First and foremost, I want to make music that I’d listen to and, ideally, music that you would listen to as well. Even though this is a very conceptual and experimental project, I don't want the resulting music to be purely conceptual or experimental. Admittedly, since I am primarily a programmer and visual artist by training, making pleasant-sounding music will be one of the more challenging aspects of this project.
  2. Make music using thoughtfully-curated data and sampled sounds - As mainstream music moves more toward computer-generated and sampled sounds, the selection and organization of sounds become the artist's narrative. This is also the primary role and practice of the modern DJ. I want to find sounds and data that interest me and mix them in a thoughtful and deliberate way.
  3. Make music that expresses data and sampled sounds in new and effective ways - By taking advantage of the temporal and emotive quality of music, I want to create new and potent experiences around data and sampled sounds that could not be achieved if they were examined independently or through another medium such as a written report or a data visualization.
  4. Make music production more accessible - I intend to document and publish the entire process of making each song from the initial idea to the production of the track. All custom software will be made open-source. This is all a learning process for me, and my learning materials and tools will mostly come from existing open source projects made available by generous people. It is only fair that I do the same.

Why data and music?

As a computer scientist and engineer, I have been conditioned to acknowledge and appreciate the impact data can have on our understanding of the world. However, as an artist, data by itself, or even popular ways of communicating it like data visualization, seems to lack a certain human or emotional component. I much prefer the traditions of storytelling and personal narratives to explore ideas and issues.

That's why the intersection of data and music is appealing to me. Since music is inherently more ambiguous and felt, mapping data to music forces the creator to think about how the data should affect the listener's mood, i.e. how should they feel about the data over a period of time. In addition, songs gets stuck in your head, so if you can embed meaningful information a song, that information may follow the listener beyond the moment of listening to the song. When was the last time an image of a chart got stuck in your head? Data visualization has the advantage of efficiently conveying an accurate representation of data. The advantage of a musical representation of data is that the creator can manipulate mood and curate a memorable narrative and experience around the data.

What are your biases?

  1. Sampled sounds over recorded or generated sounds - In the tradition of Hip-hop and DJing, I prefer to use sampled sounds as my primary instrument. This could be anything from existing music to voice recordings to ambient sounds. I like to think of it as being a documentary filmmaker rather than a fiction filmmaker, where my job is to dig through primary source material and try to figure out how to put it together in interesting and meaningful ways.
  2. Free and open over copyright - When looking for sounds, I will have a strong preference for those that are free to use, such as those in the Public Domain or Creative Commons. Otherwise, I will try to adhere to principles of Fair Use and transform existing sounds into something new. I want anyone to be able to re-use my sounds without having to get permission by the original copyright holder.
  3. Storytelling over reporting - I am interested in using data to tell a story rather than simply to report it. Data visualization already does a good job at reporting data efficiently and accurately. I would like to focus on creating experiences and narratives around that data.

What is your process?

Although my process is not set in stone, each song project loosely adheres to the following sequence of tasks:

  1. State an objective - Choose a subject that is as narrow as possible and determine specific criteria for a successful song.
  2. Research and Learn - Gather as much information about the chosen topic from primary sources.
  3. Find relevant data and sounds - Scour the internet for datasets and sounds that fit my objective.
  4. Determine a compositional strategy - Based on the objective, what is the most appropriate approach and algorithm to sequence the sounds
  5. Experiment
    1. Process the sounds to be used as instruments
    2. Process the data to generate a music sequence
    3. Organize instruments into music sequence using an algorithm
    4. Adjust settings
    5. Repeat some or all steps above until satisfied
  6. Post-process final song output - Optionally add flourish and optimize sound.
  7. Analyze and publish results - Transcribe the process of creating the song, design any relevant interactive elements, and publish to web.

What technologies do you use?

I try to use open source and free technologies whenever I can so it is easier for anyone to replicate my setup. This list will likely grow over time.

  • ChucK - a free and open source programming language for real-time sound synthesis and music creation. I primarily use this for sequencing my sounds samples programmatically.
  • Audacity - a free and open source digital audio editor and recording computer software application. I primarily use this for the pre-processing and post-processing of my sound clips.
  • Python - is a widely used general-purpose, high-level, open source programming language. I primarily use this to process data so that it can be used by my other music-related software.
  • Hydrogen - is a free drum machine application. I primarily use this to experiment with pattern-based drum programming.
  • Processing - a programming language with a visual focus. I use this to optionally generate any supporting visualizations for my songs.

What are your data sources?

I try to look for data that is reliable and publicly available, primarily those released by government or non-profit entities. I list my sources for each song, but some big ones are:

Where do you find your sounds?

Similar to my data sources, I look for samples that are in the public domain or Creative Commons. For everything else, I try to get permission from the owner when I can or do my best to adhere to the principles of Fair Use. I list my sources for each song, but some big ones are:

Who are your influences?

These are influences in terms of music style and/or philosophy. This list is actually much longer in terms of my overall taste in music, but this subset represents the most influential individuals in terms of this particular project which is grounded in hip-hop and electronic music. In no particular order:

What license is your work published under?

All songs in Data-Driven DJ are licensed under a Creative Commons Attribution 4.0 International License. You are free to:

  • Share: copy and redistribute the material in any medium or format
  • Adapt: remix, transform, and build upon the material for any purpose, even commercially.

Please provide the name (Data-Driven DJ by Brian Foo) and URL (https://datadrivendj.com) of the project where possible as attribution. Also, please shoot me an e-mail of any derivative works; I'd love to see and share them!

Who are you?

My name is Brian Foo and I am a programmer and visual artist living and working in New York City. I don't have a formal background in music production, but have always wanted to learn. My projects and profile changes, so you can find the latest on my personal website.

How can I follow your project?

You can follow along via Twitter, Facebook, Soundcloud, Vimeo, or my mailing list.

How can I learn more?

Send me an email at hello@brianfoo.com and I’d be happy to answer any questions as soon as I can.