If you want to reduce the vocals from a music track you can use this effect. You can select one of the three methods for reducing vocals.
This method utilises a machine learning model to automatically estimate vocals in a given audio. This method doesn't require the audio to be stereo. Use vocal reduction level according to your needs. Light will produce less artifacts but less vocal reduction, while aggressive will reduce vocals more but produce more artifacts and noise.
This method will attempt to identify the voice in the left-to-right spectrum of a stereo recording and remove it. The recording must be stereo (from an original stereo source like a CD - simply converting a file to stereo will not work). It will also remove any instruments near the voice in the stereo spectrum.
This option works best if the vocals are recorded in the center of the stereo recording with instruments spatially separated. If the stereo is simply a mono copied to the left and right channels, the lack of spatial separation will result in silence rather than a relative reduction in vocals.
Note: it is impossible to remove the vocals perfectly without the original mix track. You will notice some instruments might be removed too and some vocal remain. The effect will also not work on some files which have previously encoded in a highly compressed form like mp3 (because this remove some stereo depth).