Hark! The blobs sing!

Or at least, they do in Google’s latest machine learning experiment, the awe-inspiring Blob Opera, which will see a chorus of four adorable, colorful blobs serenade you with spine-tingling operatic music. Drag a blob up or down, and you’ll change what pitch they sing in; drag them from side to side, and you’ll change the vowel sound. Each blob will also harmonize with the others, in what can only be described as magical.

The Blob Opera just sounds beautiful, with soaring harmonies ringing out from each blob. Four actual opera singers — Christian Joel (tenor), Frederick Tong (bass), Joanna Gamble (mezzo‑soprano), and Olivia Doutney (soprano) — recorded 16 hours of singing (Ingunn Gyda Hrafnkelsdottir and John Holland-Avery also contributed), but it’s not their actual voices you’re hearing when the blobs sing.

Rather, the team trained a machine learning model on those voice recordings. The blobs are singing what the algorithm “thinks” opera sounds like, based on what it learned through the training. An additional model works to enable the harmonizing.

Created by David Li working in collaboration with Google’s Arts and Culture team, the Blob Opera isn’t just a cute toy — it’s a great example of how machine learning can be levered to create something new and unexpected out of existing data.

The machine learning-based nature of the opera is why the blobs are limited to vowel sounds instead of actual words, but the random noises still manage to approximate the gist of a true opera, in spirit if not lyrics.

But the best part of the Blob Opera isn’t just the lovely harmonies or adorable characters — it’s the “holiday surprise” that’s activated by clicking the Christmas tree icons, which will give you the option to let the blobs sing any number of popular Christmas carols.

Joy to the world, indeed.