6 Replies Latest reply on Aug 2, 2016 8:50 PM by Bob Howes

    How can I extract vocals from a track to use as an acapella?

    danbs10 Level 1

      Hi, just as the title suggests I would like to know how to extract vocals from a song so I can then use it as an acapella using Audition CS5?


      Also whilst I'm here, again using audition CS5, how can I determine the bpm of whatever sound file I have open, and furthermore how can I then match the bpm from another sound file to match the bpm of the track I'm working on?


      Thanks for your time



        • 1. Re: How can I extract vocals from a track to use as an acapella?
          _durin_ Adobe Employee

          Your success with isolating vocals from a song depends almost entirely on how the song was mixed, and will vary greatly.  There is no perfect solution, though many times the artifacts that remain can be minimized within a mix.  Generally, you'll have the best success with music where the vocals are panned to a specific location in the mix - usually dead center, but the tool can be used to isolate any position in the stereo field - without too much instrumentation spreading into this area.  Bass sounds are usually mixed center as well, so you may need to to some additional EQ or Spectral removal of low frequencies after you complete the extraction step.


          The tool you'll want to use in Audition is the Center Channel Extractor effect, located under the Effects > Stereo Imagery menu.  Start with the Acapella preset, then change the Extract parameter to Custom for some additional control.  Select the region you wish to isolate (it's often better to isolate passages separately as, for example, verses can be mixed very differently from a chorus) and begin playback with the effect open.  Use the Power button to temporarily bypass the effect so you can note the differences. 


          Adjust the Phase Angle and Pan parameters to maximize the vocals while minimizing the instrumentation.  Use the Frequency Range parameters to help reduce some of the low-end and high-end mix that is not associated with the vocals.  The two vertical controls on the right allow you to adjust the levels of what's being isolated vs. everything else. (Flip-Flop these to make a Karaoke mix.)  Under the Discrimination tab, the most important parameters will be the Crossover Bleed and Phase Discrimination controls.  You can adjust the FFT parameters under the Advanced tab, though I wouldn't recommend it until you're comfortable with the tool.  You don't necessarily need to know exactly what ever parameter does, but listen closely while you make adjustments and you'll begin to understand how each parameter affects the result.


          Screen shot 2012-05-02 at 10.31.27 AM.png

          • 2. Re: How can I extract vocals from a track to use as an acapella?
            samsaqib Level 1

            dear i want only voice without background music?how can i do?can u tel me plzzzz

            • 3. Re: How can I extract vocals from a track to use as an acapella?
              ryclark Level 5

              The instructions in the post above yours from Durin tells you exactly how to do it.

              • 4. Re: How can I extract vocals from a track to use as an acapella?
                Dangerpays Level 1

                Awesome! Thanks for this. It worked great for me. I followed your instructions to the T and was able to remove most of the unwanted background music from my file. Getting into the Discrimination tab was the real key for me. When I adjusted and experimented with the parameters within that tab I found I was able to greatly reduce the unwanted bg music and enhance the vocals nicely. I used your technique as well as a little bit of the noise reduction tool where there was music, but no vocals as the speaker was taking a pause. Thanks!

                • 5. Re: How can I extract vocals from a track to use as an acapella?
                  HarleyTDavis Level 1

                  Audition really doesn't work in BPM terms.  It works in samples and milliseconds.  If you have a track with BPM information, you're getting it from a DMCW not a general DAW.  There's a big difference here.  DMCW is for Digital Music Creation, and is in fact a type of DAW all it's own.  A general DAW only deals with recording live audio.  However, I believe you can use a metronome signal and synchronize the count of your metronome to the music notation.  You'll need at least 2 full lines of music (4-6bars each) and you'll have to time your metronome for that beat, then match the sound to the metronome marker by speeding it up or slowing it down to match.  Generally speaking, this is particularly difficult to do if you don't have another metronome signal.  It's best to always record one.  The easiest way is to have a drum major or the vocalist, tap the beats on a keyboard plugged into a software instrument.  IF you can see the drum major, you can keep your count from the control room by tapping it out.  You can use several different notes to indicate different counts.  If your board is midi, you can save a midi file that can be used for a metronome, and the data is much smaller.  The difference is that MIDI files aren't useable by all DAW's.  You'll need an "instrument" to play the midi, and create the sound data from the keyed notes before it can be used by Audition, and even then, if you haven't got one for both files, it's difficult to match.  And always remember, if you speed up or slow a sound, some of the imperfections can get so much worse...  Some say compression and attenuation before the speed and slow opps.  Before speed is okay.  Before slowing? No.  You'll just make the artifacts and slight noise longer.  Slow it, then apply the factor to your compression timing.  If you drop to 50% speed, double your compression attack\release, then drop slightly by 1 or 2 ms. 

                  I wonder if anybody could create a midi that senses conducting movement?

                  I've had to match the counts for some videos being played at performances with subtitles using nothing but dvd player\projection.  I did it by recording a rehearsal, matching the notes to timing, then adjusting for a broad count measure, like a quarter note in 4\4 time, and finally, counted through the music, to get the starting point of each phrase, offset by 10 frames before (preemption) and placed the title.  When I was done, It felt like the titles were spot on, even though they preempted the sound in my video mix.  I added a lead in to get the conductor in line with it, and any slight variation, natural or otherwise was undetectable.  It went over extremely well.  Just an example of trying to get the bpm to attach.  I've also done it with video that was already created.  I had to get the video to play at the speed of the song.  Similar prep.  Start by getting a count from following the sheet while listening to audio, set up a burn in counter on a video track with only the audio playing.  At the last note, or last count, stop replaying the counter, bring the last one up onto another video track, place the video on video track 1, and see how much longer it is by count mark using the upper area.  Divide upper markers by number of lower.  Add to 1.  This gives you a percentage based value to work with.  Your video is This Percentage of your audio.  Now you can adjust your video timing to fit using percentage or time\frame.  Divide the time by this number to get an exact time of the video, plus or minus some framing...  I've gotten similar results.  You can do the same with just audio.  Find the starting point, place a marker, find a decent end, use that for the size of the marker, divide the time in milliseconds by the full beats (in 4\4 there are 4 to a bar, get a general idea).  Do the same for the next file, compare the times of each songs single beat and full bar (if in the same time), or the same number of beats if they are in different time.  Speed or slow one song to the other based on that analysis

                  If song A needs to Adjust to song B, Divide the time for the same number of beats in song A by those in Song B.  Remove the decimal.  Set the speed of Song A to this percentage.  If you have a 0NN value, think 0N.N as the value.  Say song A has 1.3s for a full bar, song B 1s.  A is 130% of B for time (1.3 divide by 1 = 1.30 and remove decimal--> 130, if it came to 1.033 id go 103.3) , so I need to speed it up that much to match the timing.  Now the BPM should match fairly close. You can vary the speed by 1 or 2% to get it closer.

                  • 6. Re: How can I extract vocals from a track to use as an acapella?
                    Bob Howes Adobe Community Professional & MVP

                    Er, you did notice that this thread is 4 years old and that even the most recent post you replied to is from a year ago?