The way you use it is that the interface shows the four blobs, and you can drag them up or down to change the pitch they're singing in. You don't hear the voices of the original singers, you hear what the ML model learned opera sounds like. They recorded many hours of singing, and this was used by the machine learning model to learn what opera singing sounds like. This is the end result of a machine learning experiment by artist David Li, and a team at Google that developed a machine learning model trained on the voices of four opera singers: tenor, Christian Joel, bass Frederick Tong, mezzo-soprano Joanna Gamble and soprano Olivia Doutney. Once you get beyond the initial horror, it's an interesting (ok, and fun) idea. Google says you can be sure this will be the first blob opera, though of course, depending how cruel a set of techie friends you have, the inboxes of your contacts may already be overflowing with blob operas. You can then record your creation - and here comes the good part - share it with family & friends, especially the person who already seems to have everything. Google suggests that you can guide the pitch and vowel sound of four festive blobs who stand ready to transform your musical ideas (no matter how good or bad) into beautiful harmonies. The Blob Opera looks as bad as the idea sounds (or as wonderful, if you're on the 'isn't winter wonderful' side of the bah humbug spectrum).

Google has put its immense resources to a new use - developing the Blob Opera, a new machine learning experiment that lets you create your own festive song inspired by Opera on Google Arts & Culture.

If you're already traumatized by festive season overload, look away now. Create Festive Songs With Google's Blob Opera

Blob opera songs