Joshcore - AI generated covers of Null

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
MrB

MrB

MATI Ross
True & Honest Fan
kiwifarms.net
Joined
Feb 22, 2021

Wanna make something like this?

Joshcore is a term coined by Null for any AI cover of him. I made and posted several in the Mad at the Internet thread which were well-recieved. The amount of requests and discussion made me think that starting a dedicated Joshcore thread would be the way to go. For posting non-Jersh related AI covers you found or made, check out the general AI music thread.

The easiest way to make an AI cover is through Google Colab, a cloud-based computing service. This means that you can make these covers on Google's dime and you don't need a good PC to make covers. You'll need a Google account and some sort of audio editing software (I reccomend Audacity.)

9/18/23 edit: Google decided to ban free users for Colab after a few months so I had to rewrite the tutorial. I have no idea why the new Colab I linked to works while the old one got shut down because I'm not a tech person. It'll likely be subject to change again in the future. I have this thread Watched but I get the bug with the forum a lot where it stops telling you when a thread you're watching has new posts so if the links get fucked again just ping me. If you're a discordfag you can join the AiHub Discord Server for up to date tutorials and other voice models.

Tutorial for Dummies (9/18/23)

Tutorial rewritten. As a disclaimer, this new tool takes considerably longer to start up and constantly fucking crashes on me. I recommend looking up local install tutorials to download it on your computer if you have a modicum of technical skill since it's way faster and Google is clamping down on Colab.

This tutorial is for Windows. I have no clue if this works on a Mac or on Linux. If you use Linux you're probably autistic enough to figure it out on your own anyway.

Start by downloading the new and improved Null AI model (112mb) which is an improved version of the older Null AI model (237mb) that you don't need.

EDIT: Some other lolcow models. They can go in the other AI thread.
Ethan Ralph by me
Cyrax by @Strasse

Open this colab link.
tutorial 1.png
At the top of the page, go to Runtime > Run all. It'll take a while so just give it some time.

Once it's done, click the public URL link that appears here.
tutorial 2.png

Go to the "upload model" tab at the top of the page you just opened, then upload the Null .zip file you downloaded earlier.
tutorial 3.png

This newer UI allows you to directly link a YouTube video to any song, and it'll automatically separate the vocals, create the AI cover, then recombine the vocals in one click.

tutorial 4.png
1/2) Click "refresh models" then select Josh
3) Link your YouTube link or directly upload the song file.
4) If we have a high pitched male singer or female singer, set the pitch change to -1.
Pitch change on the top only affects the singing voice. A -1 on that setting is equivalent to a -12 on the Overall Pitch Change, which will also pitch down the instrumental as well. If 0 for pitch shift seems to high and -1 seems too low, setting the overall pitch change to -3 or -4 is usually a decent sweet spot while avoiding making the instrumental low enough to be distracting. Each number in the pitch change represents one semitone, with 12 semitones in an octave.
5) Click generate.

The quality is decent enough if you're not too invested in it or just want to make something quick. This is what I got after plugging in the theme to Baldur's Gate 3 with default settings.



At this point you'll need two audio files: the instrumental for the song you want, and the isolated vocal track for that song. Some music studios release the vocals and instrumentals themselves. In that case, just look for "studio acapellas" and download them. These will be the highest quality ones you can get. Otherwise, you'll need an AI vocal isolating tool. There are several online that are all about the same quality. Most will require an email to sign up. If you have a good GPU or are willing to wait a while for longer loading times, use Ultimate Vocal Remover. Scroll down to the main download link and use the installer that will walk you through the setup.

Now that you have your Nullified audio, open up Audacity and drag and drop your instrumental file and vocal file in. If you used AI to separate the vocals, they should be already lined up properly and in time with each other. If not, you'll need to manually line them up. If you have a bad ear for tempo and timing I can't help you there because I have good rhythm. /brag


After you've combined your instrumental and vocal tracks again, save and export them as mp3. Making it a .wav file and uploading it here will give Dear Feeder an aneurysm because .wav files are way bigger.

Here's some songs I've made so far.

Adele- Someone Like You


Chris Chan- Virgin With Rage


Russell Greer- Yo, Yovanna!


Funiculi Funicula


Leonard Coen- Hallelujah


Pharrell Williams- Happy


Icona Pop- I Love It


Blaxxy Girls- If You Feel My Love (Chaow Mix)


Bon Jovi- Livin' on a Prayer


Johnny Rebel- Quit Your Bitchin' Nigger



Red Sun in the Sky


Frank Sinatra- Send in the Clowns


Wowkie Zhang- Sunshine, Rainbow, White Pony


ASi- Super Idol



Brother Hao- Without the Communist Party There Would Be No New China


Fei Yu-ching- Yi Jian Mei


Standing on the Grassland Looking at Beijing (DJ Liu Shaofei Version)


Yakuza- Baka Mitai


The Wolfe Tones- Come Out Ye Black and Tans


Radiohead- Creep


Hava Nagila


Frank Sinatra- My Way


Eminem- Rap God


Eminem- Without Me
 
Last edited:
  • Winner
  • Like
  • Informative
Reactions: Filipenisfan420, hip hop dragon, Tropical_Paradise and 82 others
Request for North Korean Joshcore.

Pochonbo Electronic Ensemble - "Boys and girls, let's support our party!"

Pochonbo Electronic Ensemble - "Prosper the era of the Workers' Party of Korea!"

Pochonbo Electronic Ensemble - "Defending Socialism!"
 
  • Like
  • Winner
Reactions: KittyGremlin, Hongourable Madisha and NoReturn
Some additional info:

-If you feel adventurous and have a modern Nvidia GPU you can run this locally by following this guide. I can generate shit locally on my GTX 1060 this way.
-This guide will give you a few more tips and tricks on how to use it. Converting voice clips bit by bit will give you better results than converting a full vocal track.
-You can use Audacity to easily separate them with Ctrl+I, then selecting the created cutout to export it as a WAV file.
-Locally ran RVC can batch convert those clips, after which you can drop them back into Audacity and line them up with the cutouts so the timing is preserved, after which you can export it as a single long file to use for final merging.
-You should take a bit more care in what voice track you process and which parts of the processed voice track you apply. Backing vocals should be kept original while only replacing the lead ones for example.
-This method works for any voice recording. Meaning you can use it on movie quotes or even recordings of your own speech.
-Ultimate Vocal Remover should also work with other clips where you have a voice you want to separate from the rest.

As an example here's two things that I've made using all that knowledge. An AI voice replacement in a scene from Snatch:

And an AI cover of El Sonidito by Hechizeros Band, AKA the Mexican Beep Song.

In both cases I've opened the extracted vocal track in Audacity, then used the Ctrl+I shortcut to split the voice tracks, export them as separate .WAV files, processed them, dropped them back in, lined them up with the split lines, exported, and then processed the extracted vocal tracks again to mute out the processed parts, so then I could layer them all together to get a cohesive result.
 
  • Informative
  • Winner
  • Like
Reactions: Gleen Japan, BooferScooper, Laquisha Guntermensch and 15 others
You must log in or register to reply here.
Back