Google’s latest machine learning experiment is a strange but compelling opera simulator. Dubbed “Blob Opera,” the gadget puts you in control of four squishy singers who perform improvised operatic pieces on the fly.
Each doughy creature in the Blob Opera has a unique voice—bass, tenor, mezzo-soprano, and soprano—that you control by dragging the blob up and down to change pitch. Moving left or right changes the vowel sound it sings.
The singers will harmonize with the highest voice leading the harmony. So if you are controlling the soprano, the other three blobs will follow its lead. Likewise, using the tenor means that the bass will sing along while the higher voices remain silent.
David Li created the gadget in collaboration with Google Arts and Culture. Christian Joel (tenor), Frederick Tong (bass), Joanna Gamble (mezzo-soprano), and Olivia Doutney (soprano) recorded over 16 hours of singing for the project. However, it is not their voices you hear when you fiddle with the blobs. They just trained the machine learning model.
As goofy as the concept sounds, it becomes rather addicting once you start playing with it. You can even record your jam session and share it with others. I also discovered that you can still take over a blob during a recording and make it sing a solo while the rest continue the recorded harmony. There is also a little holiday surprise if you click the Christmas tree.