official AI songs thread / cloned AI vocals can be used to make final songs from someone else's demo version

[Artificial Intelligence]

don't you just train it using one of the google colab notebooks? run it for 8 hours and pay google for extra time, etc

i haven't been paying attention to the latest developments outside of following whatever's in the discord. it seems like they come up with a new notebook every other week updating the new way to do things. I was going to wait for the dude working on the aaliyah vocals to reply me on discord and see what help he needs and learn the random software surrounding getting a crisper acapella

You have to pay for collab.. if you train for free you’ll hit the limit….im learning rvc
because I can run it on my shytty 1060 pc… I’ll gonna use that Aaliyah vocals u posted and make a model myself this weeknd


This is what rvc sound like

 

[Artificial Intelligence]

Im waiting for youtube to monetize my shyt because this is the easiest lick ever

Mine says it’s copyrighted for all my vid.. don’t think I can

 

[Fani Willis fan]

wait, i didn't think you can monetize if the song is copyrighted? ...they have this manual process unless your channel has random vlogs and was cleared before to monetize, then you can post whatever

Damn i ain't uploading no more then

ima focus on the animations, I've seen some good ones.

 

[jensyao]

You have to pay for collab.. if you train for free you’ll hit the limit….im learning rvc because I can run it on my shytty 1060 pc… I’ll gonna use that Aaliyah vocals u posted and make a model myself this weeknd

where's the link to the rvc notebook? the one i saw was related to connecting it to google colab, where is the localized version?

Google Colab

colab.research.google.com

there's audacity and mvsep for free

and then you have to bootleg a whole gang of software to enhance the vocals and reduce the background instrumentals: rx 10, fl studio, and adobe podcast's voice enhancer AI

and then chop it up into various 6-10 second segments, separate the harmonies from the main vocal, separate the chorus voice from the verse voice. aaliyah's voice is dynamic, her pitch varies a lot

i'm just waiting around to see if any person can write a notebook to automate the dataset preparation process of chopping it into audio increments because it's still a lot of grunt work even after your filter out the vocals

 

[Artificial Intelligence]

where's the link to the rvc notebook? the one i saw was related to connecting it to google colab, where is the localized version?

Google Colab

colab.research.google.com

there's audacity and mvsep for free

and then you have to bootleg a whole gang of software to enhance the vocals and reduce the background instrumentals: rx 10, fl studio, and adobe podcast's voice enhancer AI

and then chop it up into various 6-10 second segments, separate the harmonies from the main vocal, separate the chorus voice from the verse voice. aaliyah's voice is dynamic, her pitch varies a lot

Are you in the ai hub discord? There’s a tutorial in there in #making models guides for running it locally
I can’t post a link I think but this is the software

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/releases/download/20230416updated/updated20230416.zip

 

[DrexlersFade]

Where that new Pun at?

 

[jensyao]

Are you in the ai hub discord? There’s a tutorial in there in #making models guides for running it locally
I can’t post a link I think but this is the software

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/releases/download/20230416updated/updated20230416.zip

yeah i saw it, looks like it's a software with a step by step guide and not a python notebook. should be straight forward, except that it's garbage in, garbage out, so the training set audio has to be pristine and that will take the majority of the time cleaning and splitting by pitch so the software doesn't get confused

i hit you up on discord, not sure if you saw. i was going to get all the people involved to divide the work and then send it to that guy who has the CPU to do it

 

[Double Burger With Cheese]

yeah i saw it, looks like it's a software with a step by step guide and not a python notebook. should be straight forward, except that it's garbage in, garbage out, so the training set audio has to be pristine and that will take the majority of the time cleaning and splitting by pitch so the software doesn't get confused

i hit you up on discord, not sure if you saw. i was going to get all the people involved to divide the work and then send it to that guy who has the CPU to do it

I got an rap artist. We got over 30 songs recorded I that I can get the session files for. Can I use this program to make a model of him and be able to type in lyrics and make a new song?

This nikka be getting on my nerves cause he so picky/indecisive on beats. He always be wanting to hear his voice on some shyt before he know he wanna do something on it for real. So this type of shyt would be perfect if I could build a model and he can just feed his lyrics into the shyt, and he can hear how his voice would sound on different beats without actually going to the studio first

 

[jensyao]

I got an rap artist. We got over 30 songs recorded I that I can get the session files for. Can I use this program to make a model of him and be able to type in lyrics and make a new song?

This nikka be getting on my nerves cause he so picky/indecisive on beats. He always be wanting to hear his voice on some shyt before he know he wanna do something on it for real. So this type of shyt would be perfect if I could build a model and he can just feed his lyrics into the shyt, and he can hear how his voice would sound on different beats without actually going to the studio first

the current model is have someone rap the demo version and then have the AI voice model clone it...

yeah you can type it out but it would sound funny, and that's another procedure to train it

☊ Nas v3 (Nasir bin Olu Dara Jones) TTS Computer AI Voice Soundboard

Type your text and hear it in the voice of Nas v3 (Nasir bin Olu Dara Jones) by justinjohn0306. 87 MP3 clips & quotes to play and download

www.101soundboards.com

 

[Double Burger With Cheese]

the current model is have someone rap the demo version and then have the AI voice model clone it...

yeah you can type it out but it would sound funny, and that's another procedure to train it

☊ Nas v3 (Nasir bin Olu Dara Jones) TTS Computer AI Voice Soundboard

Type your text and hear it in the voice of Nas v3 (Nasir bin Olu Dara Jones) by justinjohn0306. 87 MP3 clips & quotes to play and download

www.101soundboards.com

Ok got you. It’s seeming like a tool to mimic other artist more than a tool artist can use to enhance they own workflow

 

[Artificial Intelligence]

Ok got you. It’s seeming like a tool to mimic other artist more than a tool artist can use to enhance they own workflow

It can if you have ghostwriters
They send you the beat a with their voice on it .. u just have to re-record it better

 

[Double Burger With Cheese]

It can if you have ghostwriters
They send you the beat a with their voice on it .. u just have to re-record it better

Yeah that’s what I’m saying though. shyt is great for the reference tracks and I’m sure gonna be used a lot in that way. Plenty of references tracks being made for artist and writers can use it to give the artist a more clear vision on how they will sound over it.

shyt is actually gonna hurt session singers and vocalist the most it seems. They the ones doing these reference track demos. And R&B is more reference track based than hip hop

 

[Leasy]

 

[Artificial Intelligence]

Reference track :

Weeknd version:

 

[Artificial Intelligence]

converting original music is the wave! Getting so much more feedback