RVC AI Audio Generator - It's retard proof

[Black Man Underwear]

It converts any .mp3 you input into the voice of your choice, and you can download it offline so ElevenLabs can't shut your fun down. Follow these instructions to install it, ignore any suggestions to use Google services:

https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing

https://colab.research.google.com/drive/1TU-kkQWVf-PLO_hSa2QCMZS1XF5xVHqs?usp=sharing

https://docs.google.com/document/d/1KKKE7hoyGXMw-Lg0JWx16R8xz3OfxADjwEYJTqzDO1k/edit

This should take you 20 minutes maximum. Download models from the internet, Hitler, Ongezellig, and all sorts of models exist. It delivers pretty terrible results depending on your model but much easier to install than others.

Where do I find voices? I refuse to search the internet.

https://docs.google.com/spreadsheets/d/1tAUaQrEHYgRsm1Lvrnj14HFHDwJWl0Bd9x0QePewNco/edit?usp=sharing

How do I isolate instrumental/vocals in a song?

https://github.com/Anjok07/ultimatevocalremovergui/releases/tag/v5.5.0

 

[Colon capital V]

Ah another one of these threads, I was planning on revamping the Elevenlabs thread to be more of a AI Voice general; but between that, tortoisetts and playhouse, this is probably the best between them all in terms of vocal freedom of control and quality. (Tho elevenlabs probably is still the best in terms of clear and clean voice cloning, whereas RVC still sounds a bit fucky at times if your dataset isn't good enough.)

Ongezellig,

Someone from the Sharty has been posting ai covers using the voices of the 3 girls and they go unreasonably HARD between one another.
Examples below:
Mymy - I Can't Fix You (fnaf)

Maya - Die in a Fire (fnaf)

Mymy - Plastic Love (pic below rel)

Mymy - Gemmier Than You (cover based off this I think)

Coco - Don't Forget from Deltarune

Maya - I Can't Stop The Loneliness

Maya - Scott Pilgrim Ruined a Whole Generation of Women

There's also been a few edits using their voices in vids, dunno if they were made by the same chud or not.

There's also uh... this

Spoiler: toilet pussy

 

[std::string]

Alright just in case anyone wants to pop this on runpod. Use the pytorch instance.

You'll need to ssh in and install ffpmeg and libssh2-1 and you'll also need to manually install libnettle7 from Ubuntu 20,04

pip install -r requirements.txt in the directory once you've cloned it. But for some reason the version of gradio it requires is outdated, so upgrade that.

Find and replace /content/ with /workspace/ in the ipynb

Comment out all the code about google authentication and google spreadsheets because this is runpod we don't have such things here.

In whatever webUI you're running edit the .py file and change the final if to if not so gradio spins up a public instance.

I'm just going off memory but I think that covers everything.

 

[Boobalover]

Just wanted to add that the shit you make with RVC works with another program called Okada Voice Changer:

https://github.com/w-okada/voice-changer

This program will change your voice in real time using whatever model you have. This works pretty well assuming you have a good model and a good gpu.
Example of it in action (mind the anime):

By using a virtual cable you can use this to talk to people live in voice chats. We are quite close to the point where any random guy online can grab your voice and immediately start using it over his own for whatever he wants.