Joshcore - AI generated covers of Null

  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Jersh AI has been putting time in on Duolingo to cover the Karaoke classic "Linda, Linda"

The Blue Hearts - リンダリンダ
[The Blue Hearts - Linda, Linda]
Sounds great, wow. I don't know why, but foreign Null usually works better than English singing, at least for me. I end up having to sing some of them myself to avoid AI screeching and garbling etc., but I'm tone deaf, so it's a bad tradeoff. I should be looking for more resources or guides, probably just doing something wrong with RVC.



Found some old classics that I'll be working on. Also, 1k posts! Woo!

Portal (Jonathan Coulton) - Still Alive


Lazy Town - You Are A Pirate


Rednex - Cotton Eye Joe
 
Made some minor tweaks to RVC and started using the MDX-Net process instead of Demucs in UVR5 (Ultimate Vocal Remover). Seemed to help a little bit!

Futuristic Sex Robotz - Fuck the MPAA


Joe Esposito - You're The Best


The Tetris Liloet Memorial Society - Yiff in Hell Furfag
 
Some scientology songs because why not.

MC Lars - Scientology = WTF (Party at the Church of Scientology)

Nuclear Bubble Wrap - ScientLOLojyuuichi

The Fabulous Entourage - L. Ron Hubbard
 
The Irish Rovers - The Dubliners & The Pogues
Somehow AI Jersh sounds more drunk than Shane MacGowan

ABBA Money, Money, Money
Unironically could be used for MATI in regards to debanking

Gloria Gaynor I Will Survive
Jersh vs drop Kiwifarms
 
Jesus, this thread has been dead. Not that I'm contributing much, I heard this diddle as a stream outro and decided to put it through the ringer. I think the odds are some sperg here knows the actual song name and author.

?Null? - Negerbus

 
Joshcore version of Matt Rogers - Illegals In My Yard
View attachment 8335063
There was a better version of this on Youtube but to my surprise, it got deleted.

Damn it, I still have the removed version bookmarked. The rule of the innanet is that if there's the slightest chance that a piece of content can be copyright struck, archive it. I did my duty today with this other classic:

 
Considering the leaps and bounds AI-generated content has taken in 2025 alone, are there any updated tools available for creating modeled vocal tracks? Even an improved GUI would help, I'd really like to isolate few seconds of the track and quickly sample how changing some value will affect it.

Anyway, here's Wonderwall:

Kiwi Image Limited - Rise
 
Found out how to train my own Jersh voice! First iteration for me, trained off 25mins of the most recent MATI podcast. It seems to help a bit with the robot artifacting along with my new method for preparing songs for conversion. Requests are welcome once again! (some songs still suck to convert tho)

Most Racist Song Ever (Parody)
 
Back
Top Bottom