If ever there was a candidate for a new audio plugin for Final Cut Pro, this should be it. Removing vocals from music or mixed soundtracks with a few clicks.
Sadly at the moment this is just a web service, but what an amazing one and it sends the mind spinning trying to comprehend how this audio magic is achieved.
Lalal.ai allows a user to upload a music track to their website that has vocals in the mix. Then the clever AI 'source separation neural network' splits the audio into two stems, the track minus the vocals and then the vocals on their own. You then download the two stems.
If you were like me and thought that this separation would be a pretty impossible task, then head over to their website and listen to the examples. Better still, upload your own track (with vocals of course) to try out. You can split three tracks for free.
I uploaded 'This is what you came for' by Calvin Harris featuring Rihanna and after a few seconds of processing, got a preview of the track without featuring Rihanna and her vocals separated out. Both sounded good.
There is a demo video, but it shows the process, not a before and after stem separation example.
If we could program, we would digging in the API to try to wrap this technology up in to a plugin. You would still have to connect to the remote service, but being able to remove vocals from music tracks would be a great advantage when editing. Providing the music company allows you to!
But here's where it gets even more interesting for video editors. The service can also remove vocals from movie soundtracks. So with the caveat of us not actually trying this out, you should be able to isolate the speech from movies. Which would come in very handy should you want to cut trails or more mundanely, have to snick out or mute swearwords without access to a split master.
If you have tried the service, please leave a comment below, we are very keen to see how Lalal.ai works with some challenging real-world examples.
If you would like to know a bit more about the process and Cassiopeia, the neural network that powers the stem removal, then take a read of their blog post here.