Being able to isolate the voice from the rest of a song can be very useful for very different things: appreciating the naked performance, learning the different turns it takes, understanding the lyrics better, or being able to sample it in a new work without being “bothered” by the background music.
Acapella Extractor is a page that allows us to get it in a matter of seconds, and without having to pay a cent. Simply drag and drop a .WAV or .MP3 file and wait for this tool to work its magic.
However, we must take into account that the file we upload must not weigh more than 80MB or exceed ten minutes in length (limitations which are intended not to saturate the servers).
Another positive point is that we won’t have to create any users or register to use it. In addition, it is capable of isolating the voice in a matter of seconds. As its creator on Reddit explains, this website is based on a neural network specifically trained to be able to separate the voice from the rest of the elements of a song.
“It is a neural network trained on separate stems specifically for the task of separating stems. The code is also provided for 5 way splits: vocals / drums / bass / piano / other . Theoretically it is possible to train the code to distinguish other types of instruments but I believe the training data would currently not be available in large enough quantities for most instruments.“.
However, if you are looking for a tool that does the complete opposite, this programmer has also launched ‘Vocals Remover’: a tool that leaves only the instrumental part of a song.
This way, we can get two different versions of the same song: one ideal to use as karaoke, and another one to appreciate the voice without the distractions of the other instruments.