This program takes any audio/video file, extracts audio, separates vocal and instrumental tracks, applies a RVC model to vocals and re-mix them with the instrumental.
You'll need just at least one RVC model ( find some here and extract zip file ) store them in a folder, find an input file and you're OK!
- Generate from video/audio file (any format)
- Generate from youtube link
- Pitch shift the instrumental if the rvc voice has pitch shift too
- Fix pitch unconsistencies between vocal & instru when pitcshifting
- Do not run separation if output files already exist.
- Automatically find original pitch and fit the rvc model pitch ( if possible )
- Config file option to get instrumental and vocal audio file in same dir as original file
- Edit audio separator models
- Convert output file to same format as input (if audio, else use mp3)
Note: On Windows, install preferably these dependencies with Microsoft store
- python 3.10
- pip
- ffmpeg (dont forget on Windows to add the evironment variable )
- git
- Windows only C++ 14 Destktop development tools here
git clone
the repocd rvcRemix
(optional) you can create a python virtual environnement to avoid the project python libraries to interfere with the ones already present on your system
- run
python -m venv venv
linux
- then
source venv/bin/activate
windows
-
if python has not yet the permission to run scripts, run in an admin powershell window :
Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser
-
then
.\venv\Scripts\Activate.ps1
(if you're using powershell) orvenv\Scripts\activate
(if you're using cmd)
pip install -r utils/requirements.txt
- edit the file utils/config.json with a text editor and change the sections :
"modelsPath" : "a path to a folder"
with the path to the folder you put your models in (preferably each model in its subfolder)"workingDir" : "a path to a folder"
with the path to the folder where the temporary files will be put"mode" : "cpu"
with the mode to use, "cpu" or "cuda""keepTempFiles" : false
wether or not to keep intermediate temp files"copySeparatedFiles" : true
wether or not to copy separated file in the same directory as input; if existing, separation will not be done
if you created a virtual environnement
linux
- run command
source venv/bin/activate
windows
- run command
.\venv\Scripts\Activate.ps1
in a powershell
- to process a single file, run command :
python run.py "path to the audio file" keyword_of_the_rvc_model pitch(optional)
- to process multiple files, run command :
python run.py --bulk "path_to_your_bulk_file.json"
. File utils/bulk_remix.json provides an example.
Some useful ressources:
Linux, Mac, Windows (as in python)
Tested systems: ArchLinux
WTFPL.
This stuff is provided as is with no warranty at all, take your own precautions before using it.