I played around a few minutes with the various features. The voice removing was kind of impressive, though I don't know how novel that is.
I tried making some AI covers, too, which was kind of fun. For one of my tries, I submitted Nirvana's Smells Like Teen Spirit for AI voice generation to make a cover of Carola's song "Främling" (Sweden's ESC song 1983 which came in third, a very non-Nirvana-type of song). At first, I thought the voice sounded pretty much like a Swedish Kurt Cobain, then the more I listened to it, all I could hear was the Swedish artist Nordman, and it dawned upon me that they have similar voice styles. I tried lowering the pitch, and then I was certain I recognised the voice from another artist but couldn't place it. So I'm leaning towards the AI voices being trained on some not-so-unfamiliar artists rather than there being some cool AI magic happening, though I'm out of my depth here.
I tried making some AI covers, too, which was kind of fun. For one of my tries, I submitted Nirvana's Smells Like Teen Spirit for AI voice generation to make a cover of Carola's song "Främling" (Sweden's ESC song 1983 which came in third, a very non-Nirvana-type of song). At first, I thought the voice sounded pretty much like a Swedish Kurt Cobain, then the more I listened to it, all I could hear was the Swedish artist Nordman, and it dawned upon me that they have similar voice styles. I tried lowering the pitch, and then I was certain I recognised the voice from another artist but couldn't place it. So I'm leaning towards the AI voices being trained on some not-so-unfamiliar artists rather than there being some cool AI magic happening, though I'm out of my depth here.