It seems like those spooky AI images and sharpening now we are on the brink of AI voice and music generation – i.e. feeding in acapellas or instrumentals and the neural network generates new versions – with all the legal headaches that suggest. Is it owned by the computer or the original source artist? What point will it be indistinguishable? Like this video of a ‘new version of Eminem’s My Name Is made in 2021, Sure at the moment it’s rather stilted and clunky and fairly low res, but I expect that will change. We will get 2Pac resurrected in more than hologram, a Miku Hatsune writ large with the acapellas of dead pop stars.
At the moment I am spotting quite a few Facetune ‘fake older’ photos out there on Tumblr which are exceedingly Uncanny Valley, and I’m guessing at some point it will be seamless to change gender, age, race etc or generate completely new images. And the same will happen with music.
Also spoken word as well – trying this via Colab at the moment (it sounds terrible, dunno why, like badly Americanised version of me?). It will be interesting and helpful if I can patch edit my vocals with AI when I need to add things to a podcast. I suspect it won’t work that well for longer things as emotion isn’t included, but certainly part of the mashup/cutup/podcaster toolbox.
Be ready for podcasts that are entirely synthetic. BEEP BOOP!
Another system for music is OpenAI Jukebox, the results seem far more variable and nightmarish, a bit like those Deep Dream images of old. Especially when it goes all free jazz if left too long (see the Nirvana one for examples) via pomDeter for the Eminem video
Leave a Comment! Be nice….