A survey published last week suggested 97% of respondents could not spot an AI-generated song. But there are some telltale signs - if you know where to look.
Here’s a quick guide …
-
No live performances or social media presence
-
‘A mashup of rock hits in a blender’
A song with a formulaic feel - sweet but without much substance or emotional weight - can be a sign of AI, says the musician and technology speaker, as well as vocals that feel breathless.
- ‘AI hasn’t felt heartbreak yet’
“AI hasn’t felt heartbreak yet… It knows patterns,” he explains. “What makes music human is not just sound but the stories behind it.”
- Steps toward transparency
In January, the streaming platform Deezer launched an AI detection tool, followed this summer by a system which tags AI-generated music.
Compression throws me off.
That AI ‘hiss’ is ringing it my ears, but its very, very similar to YT uploads that have been re-encoded like 10 times. Which is a lot of them.
Out of curiosity, I made spectrograms of the AI song and concert that sounds ‘clean’ yet kinda noisy/compressed to me. I won’t tell you which one is which:
Source song, if you are curious: https://www.youtube.com/watch?v=MydFq0io-tQ
That song was a mistake that got generated when I pasted the wrong clipboard into Suno’s Lyrics window on my phone and accidentally submitted it. It has not been processed at all. Suno has a “Remaster” feature that when you are happy with the generated song you can give it a few automated “mix and mastering” passes to generate a cleaner and more dynamic sound.
I’ve mainly used Suno to mess about so I didn’t want to pay for the upgrade.
If there’s ever a need, you could try whatever open generation models use for post processing.