Why I Did This

I thought it would be kind of interesting to take something that was frequently changing and have it control a melody. So I set up a script to check Twitter for tweets about "Taylor Swift". These were averaging about 4 per minute. She happened to release her new single that day. Many of her fans needed to react, and I was converting their statements into melodies.

My first attempt was to match every letter to a note. I knew I wanted to have three messages being performed at the same time to create a "trio" of musicians. The problem with this approach was only a small portion of the tweet was used for the two bar loop. And there wasn't enough variety rhythmically to show it evolving. I wanted each combination of tweets to evoke a particular harmony.

So instead I matched every letter to a number, took the average number of the word and then mapped that to pitch. This approach allowed me to show more of the tweet. Another unintended result occurred, and that caught my attention.

To test what I was doing with faster-occurring tweets I switched the search term to #news. This term was about 80 per minute. I noticed the melody started to take on particular tonality. The notes F3 and A4 were appearing in all three of the voices. Without fail you would hear these pitches even as recent tweets were introduced. The starting scale is C major but F & A together imply the IV chord or the Lydian scale. The overall effect was a specific, identifiable harmony.

The reason behind it was the words "hurricane" and "texas" were present in all of the messages. Respectively the notes F3 and A4 were a part of every melody. It was very spooky to experience it as it was unfolding.

Another interesting result was when people would re-tweet the same message as news. The result would be two different instruments playing the same melody but with a different tonality.

The approach to rhythm was to decide on an algorithm. Originally I had consonants and vowels determining length, but it was too fractured. I needed something simpler so cohesion could be more apparent.

I settled on a pattern to follow and subdivide the two measures by word count. The limit was 64 words but this was much better than the four-word average I had with my first approach.

The image on the left shows how each note was divided. I am showing you the first six but this pattern carried through to 64 sub-divided beats.

Also note if you put the browser in the background the melody will remain and not update.
If you have any questions or comments, don't hesitate to contact me.
peter.ducharme@gmail.com
(617) 359-5659