UVR
Here you will learn how to Separate the Vocals and the Instrumental of songs with UVR.
Local
- Download and install Ultimate Vocal Remover (opens in a new tab).
In case you get a Windows message, just click on More information and then on Run anyway.
- Download the video of the song you want to use as a basis for creating the dataset or the AI cover:
For Youtube : copy the song link from the search bar and paste it onto Cobalt (opens in a new tab).
- 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 :wrench: 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)
- Select the audio or song as input and choose the output folder for the instrumental or final vocals.
You can change the type of wav or the MP3 bitrate in Additional settings, in Audio Format Settings.
- 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.
Before Processing - You can choose the output format format of your separate audio. - You can select if you only want the Vocals Only, Instrumental Only, the vocals with No Reverb Only, No Crowd Only, No Noise Only or with those effects. - If you wish, you can change the Segment Size and the Overlap in case of MDX23C and MDX-NET, the windows size and the aggression setting for VR Arc to your preference or leave it as it is.
- Finally, look for your processed audio in the location you selected as output.
Google Colab
-
First, we go to the UVR Colab (opens in a new tab).
-
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.
-
Run the Installation cell to start cloning the repository to the colab and mount drive.
-
In case you want to separate directly from YouTube, place the link in the audio input, otherwise leave it as it is.
Also you can download the music that you want to separate from Colab, go to the YouTube Audio Downloader for Separation cell, just place the link of the song in video_url and song name in audio_name, then run that cell. Your song will be stored in the Separar folder that you created in your Drive.
- 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)
Before Processing - You can choose the output format format of your separate audio. - If you wish, you can change the Segment Size and the Overlap in case of MDX23C and MDX-NET, the windows size and the aggression setting for VR Arc to your preference or leave it as it is.
- 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.