# UVR

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

  1. Download and install Ultimate Vocal Remover.
  1. Download the video of the song you want to use as a basis for creating the dataset or the AI cover:
  1. Open Ultimate Vocal Remover and download the models, with which you can separate the voices and the instrumental parts into separate audios. For that, you will have to click here 🔧 and we go to the Download Center section.
  • MDX-NET models
    • MDX23C-InstVoc HQ (the best to remove the instrumental and vocals)
    • UVR-MDX-NET Voc FT (to remove the instrumental and vocals)
    • UVR-MDX-NET Crowd HQ 1 (to remove crowd from vocals)
  • VR Arch models
    • UVR-DeEcho-DeReverb (to remove reverb from vocals)
    • UVR-BVE-4B_SN-44100-1 (to separate 2 or more voices singing at the same time)
    • 5_Hp-Karaoke-UVR (to remove choruses from vocals)
    • UVR-DeNoise (to remove noise from vocals)
  1. Select the audio or song as input and choose the output folder for the instrumental or final vocals.
  1. Choose Process Method select the type of models you need to use (MDX-Net or VR Arch) after that select the model and click Start Processing.
  1. Finally, look for your processed audio in the location you selected as output.
  1. First, we go to the UVR Colab.

  2. In our Google Drive, we created the folder Separar where we will put our songs to separate, and Vocales where our already separated song will be stored.

  3. Run the Installation cell to start cloning the repository to the colab and mount drive.

  4. In case you want to separate directly from YouTube, place the link in the audio input, otherwise leave it as it is.

  1. Run the cell of your choice: BS Reformer, MDX23C, MDX-NET, VR ARCH, Demucs.
  • BS Reformer
    • BS-Reformer-Viperx-1297 (the best to remove instrumental and vocals)
  • MDX23C
    • MDX23C 8KFFT instVoc HQ2 (one of the best to remove the instrumental and vocals)
    • MDX23C 8KFFT InstVoc HQ (one of the best to remove vocals and instrumental)
  • MDX-NET models
    • UVR-MDX-NET Voc FT (to remove the instrumental and vocals)
    • UVR-MDX-NET Crowd HQ 1 (to remove crowd from vocals)
  • VR Arch models
    • UVR-DeEcho-DeReverb (to remove reverb from vocals)
    • 5_Hp-Karaoke-UVR (to remove choruses from vocals)
    • UVR-DeNoise (to remove noise from vocals)
  • Demucs
    • htdemucs_ft (to remove instrumental, vocals, drums, bass and other)
    • htdemucs (to remove instrumental, vocals, drums, bass and other)
  1. Now we go to the Vocales folder, there we find our separate audios, we download the audio with the name of the song-Vocals, in case you want the instrumental is song-Instrumental.