Fun Drops with Artificial Intelligence will perform a real opera in the browser

Google Arts and Culture Lab has launched a fun experiment. Blob Opera is four operatic voices wielded by artificial intelligence. The AI ​​was trained using machine learning on the real voices of four opera singers. Christian Joel (tenor), Frederic Tong (bass), Joanna Gamble (mezzo-soprano) and Olivia Downney (soprano) recorded 16 hours of audio for this experiment. Ingunn Gida Hrafnkelsdottir and John Holland-Avery additionally provided background vocals.

For greater effect, the improvised opera is performed by a quartet of multi-colored jelly-like drops - these are the main stars, opera singers. What the user will eventually hear when casting with the cursor is actually determined by the AI. This is how the opera should sound based on what he learned from the recordings.

You won't be able to entertain yourself with the creation of a mini-opera on mobile phones - the project does not support the mobile version. Blob Opera runs in browsers on devices that support new web technologies. To enjoy the harmony of the quartet, click and drag one of the symbols up or down to change the pitch of the voice, and forward or backward to shape the drop mouth for vowel pronunciation. One or more singers can be asked to shut up by turning off the sound icon above the performer's head.

Performances can be recorded and used as you see fit. And if you click on the tree icon, artists will don holiday hats and perform their own versions of popular songs, such as Jingle Bells and Joy to the World. At the same time, you will not hear a single word in the song, the drops do not know how to use real speech and are limited only to vowels.