Hollywood bigshots gave this startup founder the idea to write music with robots – Business Insider
When an amateur filmmaker wants to soundtrack a video, they just
pick one of their favorite songs and hope it doesn’t get zapped
with a copyright takedown.
But for professional content, like the web videos from
publications that are flooding Facebook, or for commercials,
that’s not an option. You need music that sounds good, that’s
appropriate for the video, and most importantly, is affordable to
That’s what Hollywood producers kept telling Drew Silverstein
when he worked as a composer with Hollywood composers like
Christopher Lennertz. Silverstein could write
music, but having a human write music is a long and expensive
process, which means it’s simply not practical for many
He’s one of the founders of
Amper Music, a startup writing software that
writes music when given a few descriptors — “dark and epic,”
or “happy classic rock.”
After feedback from producers, he decided to hunker down and
create a “creative artificial intelligence,” as he likes to call
“I built the algorithm initially in a massive Microsoft Excel
spreadsheet. It took forever, but it worked,” Silverstein told
Now Amper Music is a three-year old startup with a handful of
employees and seed funding from Brooklyn Bridge Ventures, Two
Sigma Ventures, and Advancit Capital, with a product that
generates an original song in a web browser in seconds, while
doing the algorithmic processing in the cloud.
And the music it writes sounds good — not so good you’d want to
listen to it on purpose, but more than good enough to soundtrack
a car commercial. Silverstein says Amper’s music has passed
“blind taste tests,” and listeners usually can’t tell that
it was made by a computer.
Listen for yourself: Here’s an example of a “dark epic” song
Amper’s algorithm wrote:
And here’s a song that might be better for a reality TV
Silverstein is aware that people might see his software as a
threat to working musicians. That’s not how he approaches it —
instead, he sees machine generated music as totally different
from music written for expression.
He calls what his software writes “commodity music.” He thinks
that music valued for its end purpose, say, soundtracking a viral
video, won’t encroach on artistic space.
“Everyone at the company is a musician or audio
professional,” he explains, so his staff sees Amper as more of a
tool, for creators to use, as opposed to a push-button robot
threatening to take performers’ jobs.
The Amper web app allows a film maker or creator to set the
mood, style, and even emphasis points of a short song —
“basically, you can have the same conversation you could have
with Hans Zimmer,” a famous Hollywood composer, Silverstein says.
Press “render” and in minutes you’ll have a custom song you
can use freely. Silverstein won’t reveal too many details about
the algorithm that Amper uses to write its music, citing it as
the company’s main trade secret.
That’s because several other companies are trying to
replace stock music. Google recently revealed Magenta, a
open-source project that uses Google’s machine learning expertise
to create “compelling
art and music.” And British startup Jukedeck is working on
product as well.
Amper’s founder isn’t too worried about competition yet,
and sees it as a reflection that there’s a real demand for
custom, cheap, royalty-free music generated by creative
artificial intelligence. “Creative AI isn’t a mainstream
thing yet, but our peers and investors think it’s a big deal, and
it will be a big deal,” he says.