We developed a machine learning model trained on the voices of four opera singers in order to create an engaging experiment for everyone, regardless of musical skills. “The experiment uses a neural network trained on the voices of a bass, tenor, soprano, and mezzo-soprano to generate opera singing in real-time from simple movements on a user’s device,” a post on Google’s blog reads.This experiment pays tribute to and explores the original musical instrument: the voice.

“This experiment pays tribute to and explores the original musical instrument: the voice,” Google said in a blog post-Tuesday.

You don’t hear their actual voices in the Blob Opera experiment, but rather the Google AI model’s “understanding of what opera singing sounds like, based on what it learned from them”. Google explains that interactive designer and developer David Li created the tool by recording four opera singers tenor Christian Joel, bass Frederick Tong, mezzo-soprano Joanna Gamble, and soprano Olivia Doutney, for over 16 hours. Using machine learning, the blobs automatically harmonize with each other, and if you like what they’re signing, you can record your musical creation. With Blob Opera, you have a set of four blobs, whose pitch and vowel sounds you can control the pitch of their singing by moving them up or down, forward or back. Introducing Blob Opera, an experiment I created for Google Arts & Culture: /UYbZH8jXE2 A quartet of opera signing blobs is at your command, who have been programmed to sing together in four-part harmony.

You don’t need any musical knowledge or training-you just need a penchant for cutely weird cartoon blobs. Blob Opera is a wonderfully weird new Google AI experiment that uses machine learning to help you compose what’s essentially a Barbershop Quartet of pseudo-Minion-like cartoons, and the internet is already going wild for the fun and festive distraction.

Blob opera app