So I heard about this new website from the All in podcast that was made from ex google deepmind devs. Apparently they were annoyed at how hard it was to launch products from within google’s ai teams so left to found their own company.
Why you care: You can type in lyrics and specify the musical style etc., and it generates BANGERS. I’m not kidding, I couldn’t believe what I was hearing. Anyway, it’s free (for now) and I don’t think it’s really blown up yet, but I’m expecting it will.
About the ongoing debates with ai ethics, well, I don’t know what their policies are. I’m sure they used real artists to train the music, but if you specify to use a specfic artist it will be replaced with a more generic one in the same kind of style. So at least they’re not deep faking voices.
I’ve been messing with Suno. It’s impressive but it does some puzzling things with lyrical and melodic phrasing. It really like to add an extra line in the verse which sometimes make it feel off. I mean it’s still really interesting.
Here are a couple that came out pretty good:
Prompt:
An upbeat ska song about a cat named Percy who pretends to be brave but is really scared.
It’s doesn’t really hit the ska genre very well, but it’s interesting.
And
Prompt:
A spooky pop song about playing Call of Cthulhu with friends while drinking beer and eating pizza rolls.
Yeah. I gave Udio a spin. I think the vocalizations in Uido are much better than Suno. I like that Suno offers 2 versus and a chorus right away.
It seems like both have an “extend” feature to make the tracks longer. I’m currently trying to make my “Percy” song a little more complete by rewriting the lyrics a bit, adjusting the structure, etc.
Suno was a fun toy, but this thing… Writing lyrics and hearing it performed like this. I really worry that I will start listening to these in my spare time and it will crowd out actual music.