Tis the season when we’re never more than one singalong Messiah away from wishing we had a better voice.
The machine learning experiment takes its cues from four opera singers—soprano Olivia Doutney, mezzo-soprano Joanna Gamble, tenor Christian Joel, and bass Freddie Tong—who provided it with 16 hours of recorded material.
The result is truly an all-ages activity that’s much easier on the ears than most digital diversions.
Click and drag one of the gummy-bodied blobs up and down to change its pitch.
Pull them forwards and backwards to vary their vowel sounds.
Once all four are in position, the three you’re not actively controlling will harmonize like a heavenly host.
You can disable individual blobs’ audio to create solos, duets and trios within your composition.
Press record and you can share with the world.
The blobs don’t sing in any discernible language, but they can do legato, staccato, and shoot up to incredibly high notes with a minimum of effort. Their eyes pinwheel when they harmonize.
As Li describes to co-producer Google Arts & Culture below, it’s not the original singers’ voices we’re channeling, but rather the machine learning model’s understanding of the operatic sound.
Click the pine tree icon and the blobs will serenade you with the most-searched Christmas carols.
If you find yourself wanting more, have a go at the interactive Choir Li created for Adult Swim.
Ayun Halliday is an author, illustrator, theater maker and Chief Primatologist of the East Village Inky zine. She most recently appeared as a French Canadian bear who travels to New York City in search of food and meaning in Greg Kotis’ short film, L’Ourse. Follow her @AyunHalliday.