Find Your Inner Musician with Google’s New Differentiable Digital Signal Processing (DDSP)

Do you always hum while working or just out of boredom?

And, say this to yourself: “Gosh! I wish it to be a musical masterpiece.”

If yes, then I got amazing news for you. Now, you can turn your hum into a musical piece. Who knows? Maybe, you find a musician in you.

All thanks to Google “Tone Transfer”, now everyday random humming can turn into a melody.

Cut the chase, what do I need to do?

Just go to the Tone Transfer website from Windows, Mac or Android phone. Record 15 seconds of random hum; it can be tapping on an office desk, your voice or a musical instrument you are yet to master.

Let Machine Learning of Google turn your recorded tune into music with Violin, Flute or Saxophone.

Voila! Your readymade music is ready.

Don’t keep your expectations high from this tool, though. It may take a few trails before it gets a perfect recording. Make sure to cut the background noise and your mic works well. 

Magenta AI team of Google built “Differentiable Digital Signal Processing (DDSP)” library to use Machine Learning in art.

How is this possible?

As explained in this blog, simple elements are used to make complex signals by meticulously controlling several parameters.

Here, neural networks transform audio input into a Digital Signal Processor, which in turn, converts to other instruments.

Isn’t this super-cool?

Even you’re not a tech enthusiast, you can just make a song with your voice or

Here are the different types of inputs:

Acapella (singing)

Carnatic (singing)

Pots and pans (clanging)

Birds (chirping)

Cello (performing)

Synthesizer (riffing)

Pick your favourite one in Tone Transfer, and turn your voice into countless tunes: flute, trumpet or saxophone.

Meet you on the other side. Till then, signing off!