AI-Generated Music (+ Parodies / Covers / Original Content) - The machines are learning art. Observe...

Something interesting I found....
An AI researcher from Georgia Tech made something called "Weird a.i. Yankovic" that creates different lyrics for songs. It doesn't do the singing parts, that has to be done separately. He uploaded a parody of Beat It by Michael Jackson and got a copyright strike for it, but others have stayed up (Business Insider made an article about it). Granted, this was made in 2020 and the last edit on the Github was over a year ago. No clue if this still works, haven't played around with it yet but might be worth checking out. Could be used to make AI parody covers with other voices, unless someone else made a better version that does that with the singing parts.
 
I have made something of utter destruction level autism. As I said in the Joshcore thread; the very existence of this file disproved god.

I won't tell you what song. But it does feature Josh and 3 very special cows, with one possibly making their AI debut...



1690317795996.png
 
I'm genuinely surprised nobody's thought to do this one yet, so here goes:
Hurt - Ethan Ralph

 
I made this in honor of Josh's love for Madoka:

 
After a few more training tweeks, the Cyrax voice model is now ready for wider distribution.

I believe it strikes a good balance between realistic and palatable for the ears.

With this model you won't need to tweek the pitch and standard settings work fine without the need to fuck around to get good results. This voice model is fairly resistant to peaking issues so it can handle a fairly good range of songs and vocal ranges.

Next up I'll work on a Metokur voice Model - any other requests?

Examples:

`
 

Archivos adjuntos

I've got another Josh cover of an old outro song.

Grad Flow - Trilliam (Josh)
 

Unironically one of the best AI covers I've heard when it comes to realistic performance. The wrapping and artifacts from the old timey microphone do an enormously good job at hiding the imperfections and the song is mostly within the same frequencies of the extant OG recordings so nothing sounds too stretched or distorted, while maintaininig the original accent which is even more impressive.

This probably won't stay up so archiving just in case because in the future this will be a rare meme.
 
Atrás
Top Abajo