top of page

'Natural Machines’: How a musician taught his piano how to improvise

Move over, player pianos of the past.

Pianist and computer programmer Dan Tepfer lept into the future to put an experimental spin on playerless melodies and what he returned with is as fascinating as it is enjoyable.

New York City-based pianist and computer programmer Dan Tepfer lept into the future to put an experimental spin on playerless melodies. For the first time, he's bringing his "Natural Machines" show to Virginia Beach Saturday night. - Original Credit: Nicholas Joubard (Nicholas Joubard / HANDOUT)

For the first time, he’s bringing his “Natural Machines” show to Virginia Beach Saturday night.

Using algorithms, Tepfer taught a computer how to improvise – in real-time – alongside Tepfer’s playing.

“I play something on the piano, and that signal goes into the computer. The computer then makes decisions about how to respond, and that signal goes back to the piano and it plays by itself,” Tepfer said.

He describes the project as an exploration of the intersection between music and mechanical processes.

The idea was sparked after a trip to the Yamaha showroom in New York City, Tepfer said, but the groundwork was laid when he was a child.

Born in 1982, Tepfer grew up as personal computers become more commonplace in homes across the country. When his dad, a biologist, brought a computer home, Tepfer took advantage of the technology at his fingertips.

“I started programming by 8 or 9 years old. I just kept at it, developing games and things like that for fun,” he said.

He began piano lessons when he was 6, though he didn’t think to mingle his two passions until much later.

About eight years ago, Tepfer said he started pursuing the idea that music and machines can co-exist. His earliest ideas fell flat, he said, because resulting compositions maintained the mechanical rigidity he was trying to free himself of.

A player piano in the Yamaha showroom helped him focus his ideas and put his algorithms to practice. “Natural Machines” is the result.

And if hearing it isn’t enough, Tepfer uses the data collected via the computer to project a visual display of the music being played.

For each song, Tepfer creates a second algorithm to illustrate what the music looks like.

“For each one of the algorithms, I’m trying to show the structure of the music as clearly as possible and the mood, to a certain extent. For example, that ‘Inversion’ algorithm that I use – the way that rule works is that the computer responds in this mirror image to what I’m playing. I wanted to make that really obvious on the screen. I want to call on people to see that immediately," Tepfer said.

Even though he uses a determined algorithm for each song on the project, no performance will ever be identical.

“The rules stay the same from performance to performance, but there’s a lot of freedom in there,” Tepfer said.

Even though there are rules, the spontaneity of each performance relies on Tepfer’s ability to change his melodies up.

The second melody might be a result of a computer, but Tepfer has to treat it like another human.

“What I’ve learned is that algorithm aside, I need to be listening to what the computer is playing just as deeply and with as much attention as I do a human partner. I have to treat the computer as a legitimate partner.”


bottom of page