Removing vocals from music… Is it possible?

I have been asked whether I can do this a number of times over the years. I understand that people often want to do this in order to use the music as a backing track.

Now, I hate to spoil the party but this is not something that can easily be done. Well, it can, but it never really yields worthwhile results. Let me explain why.

Mixed audio

Quite simply, when you are listening to a fully mixed track that is exactly what it is. It is the result of mixing several layers of audio together including drums, guitars, bass guitar, vocals, strings etc.

These individual layers exist in a multi-track system where each track can be processed individually. This would be the ideal time for creating a karaoke version of the song by muting the vocal track. However, only the record label will have access to these multi-layered parts. We are given the two channel stereo ‘mixed’ version.

With the stereo version it is near to impossible to separate and remove individual elements. Although there are ways, they do not produce great results. Without getting too deep into the subject of frequencies and harmonics, sounds occupy their own sonic space.

For example, a piano and a voice have their own sonic range made up of frequencies and harmonics, but many of them overlap. Therefore, when they are combined to create a song they create a more complex sonic range that is impossible to separate back into its component parts.

What are the options?

If you are just looking to remove vocals there are some options, but far from ideal. If you are looking to go a step further and isolate individual elements, maybe to create a DJ style mashup there are also options. Again the results are not exactly spectacular.

Karaoke Tracks

Personally I would suggest buying the specially created karaoke versions, but if you do want to try and remove the vocals there are some software options. I have never used them but know they are available.

The big problem with these is that they utilise an old audio technique using phase reversal to remove the centre channel. In most cases this works to some extent because the vocals are usually in the centre of the mix. The problem with this is that any other instruments that are in the centre will be removed too, typically bass and some drums. Results will vary from piece to piece but overall the result will not be satisfactory in my opinion.

The free way of doing the same thing is to use phase cancellation on audacity software. I will do a tutorial on this in a separate post.

Stripping the elements of a track

The reason I have explored the following software is because we have recently been commissioned to create some mashups. A mashup is basically combining parts of two songs or more to create a new mix.

Now, while I like a challenge, using the vocals from one track and combining with a new one poses several issues. The main one being that I do not have access to the individual layers and need to somehow strip the parts I wish to use.

To the best of my knowledge there are only two pieces of software capable of doing this, however, for the reasons mentioned earlier, the results are sketchy at best.

Hit n Mix

This has a free trial version and it is a cleaver piece of software that manages to strip a track (broadly speaking) into its component parts.

From my experimentation, I found this software reasonably effective at allowing you to alter items such as pitch, but as a serious audio tool I am sorry to say the quality of the files is somewhat degraded. Also, from the point of view of stripping out vocals for use in a mashup the software does not do it well enough in my opinion. As soon as you remove a clip from the context of the original track it sounds terrible in most cases and is unusable.

I would suggest playing around with the free version before investing the £80 for the software.

Melodyne Editor

This too has a trial version which is fully functional for 30 days. I am glad of this as I soon worked out that it would not work for my mashup requirements.

The paid version is about £299 and is a serious audio production tool, used to better effect as an audio processing tool for producers working with vocal and instrumental recordings. In the context of stripping tracks it suffered with similar issues to the Hit n Mix software in that changes were acceptable until individual parts were played on their own.

While not suited for my needs, I feel that Melodyne would be a brilliant tool for manipulating tracks if you had the multi-layered versions, which brings us back to the original problem.

Izotope RX

I have and use several Izotope products including RX. I am not claiming that it can separate the instruments from stereo tracks, but it is quite effective at allowing you to visually edit the wave form using its spectrograph view. This allows you to identify and remove certain elements of a sound.

For example, it may be possible to see and delete a bass line. Although, where sounds overlap and cannot be isolated it is impossible to remove them.

This is a serious audio tool priced at £260, but there is a trial version.

To Roundup

While there is definitely a need for this type of tool among DJ’s and producers and clearly some attempts to address it, I think that the task may not be possible given the complexity of sound. Some tools may be able to help depending on what you are trying to achieve, but overall the problem does not have an adequate solution. When or if technology develops enough to provide an answer I will be one of the first keen to try it, but until then I can only hope that major record labels start to release music in layer form… fat chance!

If you disagree, have an opinion, had similar issues or found a great way round this please leave your comments.