<iframe src="//www.googletagmanager.com/ns.html?id=GTM-PWMWG4" height="0" width="0" style="display:none;visibility:hidden">

It’s deepfake karaoke, Old Master style

15 March 2021

Introducing Rakewell, Apollo’s wandering eye on the art world. Look out for regular posts taking a rakish perspective on art and museum stories.

A new tool goes a step further than those AI animations of portrait paintings that have been doing the rounds: the recently launched Wombo.Ai app, which transforms still images into singing videos. By the end of last week, some 15 million Wombo videos had been created – among them, of course, some delightfully expressive footage of historical artworks serenading the internet.

Getting things off to a disconcerting start, here’s Henry VIII getting a little too into his rendition of Ding Ding Song, the Europop parody by Swedish musical artist Günther. Click through to the Twitter thread to watch his six wives take on a range of other lively tunes.

The Mona Lisa, rightly or wrongly, is feeling herself (as the kids say):

Seventeenth-century portraiture, it seems, is especially well suited to a whole range of karaoke classics – witness Van Dyck’s Charles I singing along to Gloria Gaynor’s I Will Survive (if only he knew).

Art dealer Philip Mould has an astute observation on what makes these particular deepfakes so uncanny – ‘the […] addition of a set of teeth!’. As Mould points out, ‘formal, society portraiture rarely if ever allowed them to be seen’ until late in the 20th century. 

Wondering about the technology that makes this magic happen? Well, apparently the developers of the app recorded a series of lip-syncing videos using live performers in their studio, which are mapped on to the images that users upload. As for the inspiration behind the whole project? Speaking to the Verge, Wombo CEO Ben-Zion Benkhin revealed he came up with the idea for the app ‘while smoking a joint with my roommate on the roof’. Makes sense.