Applio
GuidesAudio isolating

KaraFan

Here you will learn how to Separate the Vocals and the Instrumental of songs with KaraFan.

Local

How to install?

Important its is only for NVIDIA GPUs

  1. On your PC where you wish to store the KaraFan project and Go inside it.

  2. First, we have to download and install Python.

  3. Download:

If you have installed Python 3.11 before WITHOUT using this Install.bat, you must uninstall it before! Even those installed with the Microsoft Store!

Important On Windows 11, you must allow first the execution of the BAT file:

  • Right-click on the file
  • Choose « Properties »
  • In the « General » tab, click on the « Unlock » button
  • Click on the « OK » button
  1. You must execute it with Administrator Rights.

  2. Run Setup.py by double-clicking on it.

  3. Now, you can launch KaraFan.pyw by double-clicking on it.

How to use?

  1. Place your audio in the Music folder and copy the path of your song that you want to separate and the path of the folder where the separated song will be stored.

  2. Now place the settings you want to apply, try the ones you want the most, but if you want to effectively separate the instrumental or sounds from the vocals, I recommend the following settings:

To separate the instrumental from the vocalsTo separate other typo of audios
Karafan VocalsKarafan Other
  1. We click on Start and it will begin to separate the chosen song. At the end of the process, the message Processing DONE will appear at the end of the interface.

  2. Now we go to the Results folder and we will find a folder with the name of our song or audio. There we will find a list of audios, we download the audio with the name Vocal Final, in case you want the instrumental one that says Music Final.

How to update?

Simply Run the Setup.py file!

For more information, visit the GitHub repository.

Google Colab

Enter the KaraFan interface

In our Google Drive, we created the folder Music where we will put our songs to separate, and Vocals where our already separated song will be stored

The Colab only lasts 30 minutes in execution. Additionally, it is possible that Colab may disconnect after a few hours due to Google’s limitations.

First, we go to the Google Colab.

Run the cell by clicking on the white play icon button. So that the repository can start to be cloned.

Then we click on the next cell This is it! to load the KaraFan interface.

Process your audio

Now we have to copy the path of your song that you want to separate and the path of the folder where the separated song will be stored.

Now place the settings you want to apply, try the ones you want the most, but if you want to effectively separate the instrumental or sounds from the vocals, I recommend the following settings:

To separate the instrumental from the vocalsTo separate other typoe of audios
Karafan VocalsKarafan Other

We click on Start and it will begin to separate the chosen song. At the end of the process, the message Processing DONE will appear at the end of the interface.

The first time you use it, it will start downloading the models you have selected in your drive in the KaraFan_user folder. If you keep this folder for the next time you want to use it, you will not need to download it again.

Now we go to the Vocals folder and we will find a folder with the name of our song or audio. There we will find a list of audios, we download the audio with the name Vocal Final, in case you want the instrumental one that says Music Final.

Code to be afk

To run this code in Colab, we have to press the keys Ctrl + Shift + i to open the developer tools option and the following window will open. In case another window is displayed, we can locate it by selecting the Console option, so in this way Colab will not disconnect during the separation process. First, we put Allow pasting, then we press enter and only then it will let us enter the code.

function ClickConnect() {
  var iconElement = document.getElementById("toggle-header-button");
  if (iconElement) {
    var clickEvent = new MouseEvent("click", {
      bubbles: true,
      cancelable: true,
      view: window,
    });
    iconElement.dispatchEvent(clickEvent);
  }
}
setInterval(ClickConnect, 6000);

On this page