I am using Windows 7. I have downloaded Audacity 2.2.2. I want to remove singer’s voice from a video so that it becomes a karaoke. I could easily upload the mp3 file through File-Open. I notice one track is visible. Not sure how to proceed further
I notice one track is visible.
You need stereo (left & right). Vocal removal works by subtracting left & right to remove the “center channel”. If the MP3 is stereo you can try it.
Vocal removal is usually imperfect. It will also remove centered lead instruments and any non-centered vocals will remain. I consider vocal removal to be a “novelty effect”, not something for serious audio production. Real Karaoke recordings are usually made from a different performance by a different artist with the vocals simply left-out, or on rare occasions they may be made from the original multi-track recordings.
Other than isolating the left or right channel, or this center-channel “trick” you cannot isolate vocals or instruments or un-mix audio. “You can’t in-bake an egg or un-fry and egg and you cannot un-mix audio.”
Note that ALL CDs are 2-channels and if you have a mono CD (with two identical channels) vocal removal will give you a totally silent file. MP3s ripped from a mono CD will usually be the same with two identical channels.
Vocal removal is a rather touchy (and tricky!) subject to discuss since lots of people say it can’t be done; To that, I somewhat disagree. More and more programs which attempt to do this stuff are showing up, and at least a few of them aren’t using any stereo trickery (Audionamix’s Mac-only products come to mind right away). There’s a web service which does both vocal removal and isolation, but it comes at $4 per song. It’s called PhonicMind. Depending on the song, it may or may not give a usable result for karaoke. It works on mono AND stereo files from what I can tell. It uses a neural network / deep learning to accomplish the task so it’ll only get better with time!
Sadly, for some songs it will just be near impossible to remove the vocals effectively… The biggest offenders are drumless piano pieces with a singer which has a load of reverb on the voice. It is just about impossible to cleanly take out the vocals and remove the reverb as well without hearing any remnants /artifacts.