Tutorial - Vocal Removal and Isolation
ToDo-2 ready for editorial review and merciless editing.
- For popular songs, consider downloading vocals-only tracks from the internet.Many studios release the instrumental tracks (with and without backup vocals) for use with things like karaoke and some singles and records even have them on the B side.
Case 1: Vocals in the middle, instruments spread round them
If the vocals are panned in the center of a stereo track (fairly common in "pop" music tracks), the so-called "vocal removal" technique can sometimes be effective by removing what is common to both tracks (i.e. the vocals), leaving behind what is different (i.e. the instrumentals).
To try this technique in Audacity, split the stereo track
"...of one of them, and play back the result; see some better text of the Audacity FAQ for step-by-step instructions on this technique, or a..." this is the non-http version of the FAQ link: some better text
into its left and right channels, make both mono, invert all (or a selected part) of one of them, and play back the result. See this page of the Audacity FAQ for step-by-step instructions on this technique, or a YouTube video tutorial covering the same process. Note this removes everything panned in the center, not just vocals. In pop music this could mean removing bass or rhythm parts. Removal of the vocals can often be incomplete and will leave artifacts behind (especially where there are backing vocals and where reverb (echo) has been applied) as this spreads sound sources, and makes them very hard to extract from each other.
Vocal Remover (for center-panned vocals)
In the current version of Audacity a Nyquist plug-in effect is included to automate the steps involved in Case 1; for fuller details see Vocal Remover. It is accessed from . It includes a Help screen and an option to retain a specified range of frequencies (so removing everything outside that range); there are three choices of removal methods that can be used, for details see here.
Although two channels of output are produced the result will be mono because both channels will be identical and panned to center.
Case 2: Vocals in one channel, everything else in the other
If you have an unusual stereo track where the vocals are mixed hard into one channel and everything else hard into the other channel, you can simply split the stereo track into left and right and delete the vocal channel. To isolate the vocals rather than remove them, delete the other channel. To split a stereo track in Audacity, click on the downward pointing arrow at the top of the Track Control Panel (where the mute/solo buttons are) then on .
To delete one of the channels, click the to left of the downward pointing arrow. If you accidentally delete the wrong channel, use to get it back. Finally, click the downward pointing arrow again and choose "Mono", so that the track will play out of both speakers.
Case 3: Vocal Isolation
If you can make a two-channel track with center vocals removed (as in Case 1 above) it is tempting to think that those vocals be isolated by inverting this track against the original track, so as to remove the non-vocals. Unfortunately, using Audacity, vocals often cannot be isolated in this way. The result of the Audacity Vocal Removal effect is a mono mix of sound that was in the left and right channels but was not common in both. Mixing this back with the original track (either inverted or not inverted) will produce a stereo track that contains a new mix of the center panned vocal and the non-center sounds. Different techniques used by some third party plug-ins such as 'ExtraBoy' claim to be able to isolate vocals given suitable audio material (see below).
Using Audacity for vocal isolation
An Audacity user posted this technique for using Audacity to attempt to isolate vocales; it will not work in all cases.
I discovered a way to get reasonably good isolated vocals. Also it is sometimes possible to isolate vocals by using Audacity's Noise Removal to capture the noise profile of a song that has had vocals removed, then run Noise Removal with that profile on the original mix before vocals were removed. See:
First, using http://wiki.audacityteam.org/index.php?title=Vocal_Removal
remove the vocals from a song. Using the song with vocals removed, sample it as noise profile with the noise removal effect. Apply noise removal to the original audio.
It's still tricky to get rid of anything else in the center (this almost always includes the bass guitar and bass drum), but careful use of EQ can get you a pretty good result.
The following may let you remove drums which are not in center but keep the vocals
If you need to retain center-panned vocals, you can try Noise Removal on the vocal-removed track to isolate the vocals: http://forum.audacityteam.org/viewtopic.php?f=28&t=4956&p=136527#p136527
Then import the original song into Audacity again and try this Nyquist plug-in to move the bass frequencies to center: http://forum.audacityteam.org/download/file.php?id=3065
Then remove the drums from the center by inversion, and finally mix the vocals back in.
Gale 16Apr12: I have just dump-pasted below raw jottings I have kept on isolation including using Brainworx, Audacity Noise Removal and inversion against a track that has just the instrumentals. Someone should be able to work on this.
- Gale: Any detailed text for Brainworx for isolation will probably want moving to Wiki too, also noting there that it is for Windows as well as Mac.
- Ed 17Apr12: did some editing but given the "notes" nature of the text I am going to hold off on any more work as I cannot really see how to "fix" some things like sentence fragments.
- But if you can remove the vocals, doesn't that imply that you can isolate them as well? Once you have the two tracks, left and right, of just instruments, couldn't you invert them and add them to the original left and right tracks, leaving just the vocals?
No.
(quite tempted to leave it at that) [Center pan isolation can't be done using the inversion, a.k.a. destructive interference, method of center pan removal; another method, spectral subtraction, is required, which is what the Extraboy and Kn0ck0ut plug-ins use]. People are told that they can "remove" center-panned vocals (through inversion of one channel then summing the channels), and then think they can isolate the vocals by subtracting that "non-vocal" version from the original.
Left audio = L Right audio = R Center audio = 2C
Left channel = L+C Right channel = R+C
Remove center audio: (L+C)-(R+C)=L-R (mono)
From the "left channel" and "right channel" (which you start with), there is no simple way to isolate C.
Subtracting the center pan removed mono from the original stereo in an attempt at center pan isolation gives …
Left channel (L+C) and Right channel (R+C) minus center pan removed mono (L-R) => (L+C) + (R+C) - (L- R) => L + C + R + C - L + R {double negative gives "+R"} => 2C + 2R
center isolated audio is “2C“, so “2C+2R” is not center audio.
If you define “Remove center audio“ as : (R+C)-(L+C) =R-L (mono) => 2C + 2L, (which still not “2C” on its own).
- For popular songs, consider downloading vocals-only tracks from the internet
- subtracting an instrumental from a song to isolate vocals http://www.youtube.com/watch?v=BKodcXMQHZU
- Brainworx Solo I use mid-side processors - usually a lead vocal will be panned center, so If I solo just the "mid" channel and mute the "sides" of the sound, a lot of extra stuff disappears (anything that was panned all the way left or right). The sides usually have the lead vocal's reverb on them too, so it helps to pull the FX of the vocal to mute the sides out. I use the brainworx solo plugin to solo out the mid (VST, au, rtas)
The "heart" of our M/S tools are our unique SOLO BUTTONS that allow you to actually LISTEN to all components of a stereo mix / stereo signal individually.
L(LEFT channel), R(RIGHT channel), M(MONO sum, mid signal) & S(STEREO-Difference, side signal) - and all this phase-corrected (when needed) and mono-ed out on both speakers of your system. This is very helpful when working in M/S modes (because especially listening to only the S-signal of a stereo mix is not so easy to do with modern DAWs) and to check your signals / mixes for errors, noises, clicks, whatever...
- Re: Extract a vocal from a song AND KEEP THE VOCAL
I discovered a way to get reasonably good isolated vocals. Also it is sometimes possible to isolate vocals by using Audacity's Noise Removal to capture the noise profile of a song that has had vocals removed, then run Noise Removal with that profile on the original mix before vocals were removed. See:
First, using http://wiki.audacityteam.org/index.php?title=Vocal_Removal
remove the vocals from a song. Using the song with vocals removed, sample it as noise profile with the noise removal effect. Apply noise removal to the original audio.
It's still tricky to get rid of anything else in the center (this almost always includes the bass guitar and bass drum), but careful use of EQ can get you a pretty good result.
The following may let you remove drums which are not in center but keep the vocals
If you need to retain center-panned vocals, you can try Noise Removal on the vocal-removed track to isolate the vocals: http://forum.audacityteam.org/viewtopic.php?f=28&t=4956&p=136527#p136527
Then import the original song into Audacity again and try this Nyquist plug-in to move the bass frequencies to center: http://forum.audacityteam.org/download/file.php?id=3065
Then remove the drums from the center by inversion, and finally mix the vocals back in.
- How about using "isolate" in Noise Removal?
- isolating the vocals works like removing vocals; in both cases we combine the original waveform with an inverted waveform to “subtract” the part we don’t want. It will leave us, in this case, with the vocal track. In order for this to work, however, you need to have a studio version of the instrumental track. Removing the vocals to get an instrumental track and then trying to isolate the vocals does not work in this case. Don’t be disheartened if you don’t have one ready. Many studios release the instrumental tracks (with and without backup vocals) for use with things like karaoke. There are plenty of places online where you can buy these tracks (like Karaoke-Version.com and some singles and records even have them on the B side) as opposed to the final mono track you get from removing vocals, this method will leave you with a full stereo track. As such, it becomes more important to try and match the quality of both tracks and align them before you isolate the vocals.Open up Audacity and import both the regular and instrumental tracks.
Select the Time Shift tool to roughly align the two properly. Next, zoom in really close and then zoom in more. Take the proper time to align this as closely as you can; pick a peak or trough in the left channel of one track and match it precisely with the left channel of the other track. If the alignment isn’t right the process won’t really work.
Invert the instrumental; Ctrl+A to select all of both tracks. Go to Tracks > Mix and Render.
You’ll get one combined track that should have a more diminished amplitude where the vocals were kept and the instrumentation removed.
Plug-ins
As well as the above methods supported by Audacity itself there are various third-party plug-ins that can be used to try and remove or isolate vocals. Please see this page in the Audacity Wiki for details.
Windows Plug-ins
Two Windows VST plug-ins are known of that can be used in Audacity for vocal removal and isolation.
- kn0ck0ut (free) can sometimes remove vocals where they are not center-panned but rather different in frequency make-up compared to the non-vocal parts of the track. It works on two mono tracks (split from a stereo track and made mono as above) by extracting the right channel frequencies from the left hand frequencies, leaving the "result" in the former left channel. Alternatively, if your vocals are center-panned, this plug-in can often make a good job of isolating them from everything else.
- Voicetrap is a commercial plug-in featuring center channel removal by frequency-based and cepstral methods (that is, not a simple subtraction of one channel from the other). It has an advantage over kn0ck0ut of providing stereo output for the "vocals removed" track, as well as offering isolation of center-panned content. There is a demo version with a "mild" vocal removal setting (but no isolation ability) that works in Audacity.
To try either of these VST plug-ins in Audacity, place the unzipped .dll file from the plug-in's zip folder, and the unzipped vst-bridge.dll from the Audacity VST Enabler into Audacity's Plug-Ins folder. The Plug-Ins folder will be in the folder where you installed Audacity, typically C:\Program_Files\Audacity. On restarting Audacity the new plug-in will appear in the Effect Menu underneath the divider. Should vst-bridge.dll not work, try the previous VST Enabler for Windows (place the already unzipped vst_Enabler.dll in Audacity's Plug-Ins folder along with the plug-in's own .dll, then exit and restart Audacity). Note that VST plug-ins in Audacity only have a generic tabular interface due to licensing restrictions imposed by Steinberg.
There is another commercial VST plug-in called ExtraBoy which offers useful functionality for those interested in vocal removal or modification of the different components of an audio track. There are two versions of this plug-in but, unfortunately, neither work properly with the currently limited VST support that Audacity can offer. It seems that the processed audio is always completely silenced irrespective of the plug-in settings used. Both versions of ExtraBoy do function with other audio editors which offer full VST support, so you could always export a track from your Audacity Project as a WAV or AIFF file and process it with Extraboy using a different audio editor.
The "lite" version of ExtraBoy is quite similar to VoiceTrap but offers some vocal removal ability on all tracks (not just on center-panned vocals). It has two "vocal removal" algorithms: (1) on the basis of the vocals' frequency characteristics and (2) on the basis of their spread in the stereo spectrum. The two algorithms can be combined to obtain the best possible removal in a particular case, and full stereo information is preserved in the processed track. A "vocal isolation" algorithm is also provided.
The 14 day full-featured "demo" version claims to be capable of isolating, removing, suppressing or enhancing any component of a stereo track, based on its identified frequency and spatial "signature". Naturally the author points out that the exact results are subject to the particular characteristics of each track. Multiple components of a track can be processed simultaneously (for example, to isolate piano and vocal, or enhance piano and bass).
OS X Plug-ins
On OS X you can try the voxReducer Audio Units plug-in (14 day free evaluation) in Audacity, as long as you use the current version of Audacity. This requires OS X 10.4 or higher, although a legacy 1.3.3 version for OS X 10.3 is still available.
VoxReducer is similar to the Nyquist Center Pan Remover in that it aims to reduce the strength of center-panned vocals, but differs in having a phase offset adjustment and a slider for adjusting the intensity of the vocal reduction.
Try putting the voxReducer.component in either of these two locations:
- /Library/Audio/Plug-ins/LADSPA or
- ~/Library/Audio/Plug-ins/LADSPA
or (as suggested by the program documentation):
- /Library/Audio/Plug-ins/Components or
- ~/Library/Audio/Plug-ins/Components
brainworx makes a vst plugin bx_solo that lets you solo mid and side channels; it also has a mono-izing knob and the ability to solo left or right channels. It also has an AU and RTAS plugin and they're free.
