"Reverse Karaoke"/How To Remove Music, Keep Vocals?

Firstly, thank you for a great application.

As an amateur guitarist I often find it necessary to transcribe lyrics from recorded music. However, it is sometimes difficult to make out the vocal lyrics due to the volume of the music.

Is there an “effect” that mutes the music while isolating and retaining the vocals , sort of like a “reverse karaoke” effect?

That’s called “Vocal Isolation.” That’s your Google search term. You can try searching the forum, but I don’t think we ever got any further than “this works badly.”

You would think if we could do Vocal Removal, we could turn the program over and do isolation. The karaoke tools work from a stereo show and the result is mono. So the shows before and after vocal isolation are different. That means you can’t reverse the process with a simple program.


See http://manual.audacityteam.org/o/man/tutorial_vocal_removal_and_isolation.html.


Many thanks for the link, Gale. The result is a bit distorted, but it does make the Vocals a bit louder than the original.

Hi Frank

Do you mean something like this?


That’s actually pretty good, better result than I managed to get.

With luck and a following wind we might see a new effect in the next release of Audacity called “Vocal Reduction and Isolation” - you can guess what the second part of that tries to do :wink:


Yes and Robert who replied wrote that effect.


Excellent!! Any chance that I could be notified when this effect is released?

We can’t guarantee to do that but if you go to the bottom of http://web.audacityteam.org/ you can sign up for our announcements. If the new vocal remover/isolator is included, that will be mentioned in the Release Notes linked to in that announcement.

You can always download Robert’s plugin and add it to the Plug-Ins folder in Audacity. Here it is http://forum.audacityteam.org/download/file.php?id=12536.


Thank you Robert for the plugin, and thank you Gale for the link to it.

Easy to use and works well. Now I need to play around with it some more.


some feedback after you have played with it - in particular what settings you find most useful and effective - would be most helpful.

I’ve been playing with it in the name of testing for Robert (usually as I record new stuff). So far I seem to have more success with the vocal removal than the vocal isolation - but I’m no expert and I haven’t had much time to play with the parameter settings.

And, like you, I’m finding it pretty easy to use - documenting its intricacies for the Manual may prove a little harder I’m thinking.


Isolation needs normally a higher strength value than one.
Here is an example with different strength settings applied:

The higher the strength the quieter the audio will be (unless you have a perfect panned sample). Thus, you’ll need to normalize the audio afterwards, 50 dB difference are not unusual for the highest setting. This results necessarily in more artefacts (quantization noise).
However, it will be masked when re-mixed.
The Remove option does not need the normalization with a high strength setting.

Back to the sample file:
The stereo version is followed by the mono mix, this is the exact counterpart to the current vocal removal. It attenuates sounds that are only present in one speaker by 6 dB while keeping the center equally loud.
The next strength setting is 0.02 (= 1/50).
The voice gets a “mini-radio” like character.
Afterwards, the strength is 1 and more. The extreme points (sound in one speaker only) are in each case maximally attenuated (~-120 dB). With increasing strength, the center gets narrower and narrower.
This makes in some cases special effects audible–chorus, flanger etc. because the center makes actually a snake-like movement and not a straight line. You’ll hear pumping.

Peter, I’d be grateful if you could work with me on the documentation.
I’ve imported the current “Vocal Remover” page and made some alterations to it.
We’ll need probably some additional tutorials, examples, case studies or however they should be named.

Are sound clips yet supported? They might be bettr than graphics alone.



Robert I’ll be happy to do that.

Have you got an editor account for the Manual? If so then you can place your revised page back there under a different name to Vocal Remover of course and then we can work on it.

Are you hoping to develop tutorial(s) in time for 2.1.1 - or are we just aiming for the basic documentation by then.


No, embedded audio is not yet supported.

If you just write something like this in the wiki page, having uploaded the file:


then if you click the “Drones” link the file will play in its own page in Firefox and Chrome and you’ll press Back to go back to the wiki page. Internet Explorer users will just be offered to download or open the file even if they have the Xiph.org codecs installed.

Otherwise (I am no expert) I think you need to install the Scribuntu extension with {{Template:Listen}} and {{Module:Listen}} ( http://stackoverflow.com/questions/24116088/install-listen-template-on-mediawiki?rq=1 ) and/or or install an audio player extension.

Even with the Xiph.org codecs installed, trying to play embedded files on http://en.wikipedia.org/wiki/Template:Listen in Internet Explorer is hard. Java is needed and you must configure Java to add the page to a list of sites that are allowed to run Java applications.

I don’t know about making any of the embedded player solutions work in the dumped Manual bundled with the release.



is perhaps not terrible. Even without the Xiph.org codecs, IE users can open the file if they have a player such as VLC that supports OGG.