Has anybody had experience of removing location music from documentary dialogue? A filmmaker has approached me to tackle an issue with broadcast rights on the music coming from a stereo in the background during a couple of shots in her documentary. I was thinking of editing the dialogue as I would for a scripted film and then replacing the gaps with ambience from the shoot as well as a royalty-free music track of similar style fitting the reverb of the warehouse where the film is shot. I'm not sure how well this will work. Is this going to be insanely difficult? Any one else had a similar job in the past? Thanks Phil

  • Yeah folks, I know. Please re-read my post. I suggested to TRY OUT phase cancellation tricks. Btw. the level is not as important as the frequency domain.
    – Larry
    Feb 1, 2012 at 8:43
  • I've had a documentary maker approach me with the same exact task. It's a mono file. I tried phase cancellation tricks with the original music file, Izotope RX, Spectral Layers (after going through all its instructional videos), creative editing, but nothing is giving me a half decent result. So I think I will have to convince him to get the characters in here and rerecord the parts he needs.
    – Larry
    Feb 3, 2014 at 19:17

7 Answers 7


I'd say insanely difficult, if not impossible. You're going to have huge issues if the music is audible while the person is speaking. If that's the case, you're going to need something that can separate out elements of a complex signal. The only thing I can think of that might be able to pull that off is Melodyne...but I'm not sure even that will work.

Is a reshoot of the interview is off the table? Because that's the filmmaker's best option.

  • Yep re-shoot out of the question. It was a one-off interview that won't happen again. Your right, its going to be pretty well impossible having listening back to it a good few times now. Its a shame as the dialogue itself is perfectly usable. Thanks for that Shaun...
    – Phil Lee
    Jan 31, 2012 at 15:10
  • @Phil Lee - check out melodyne then (i think you can demo it at least). it let's people break out and edit mixed music into component timbres and notes. it might be able to get you a little further along than you would otherwise. Jan 31, 2012 at 17:47

I would say removing the music is impossible. Phase cancellation won't work unfortunately, the signals need to be mathematically identical for this to work, and if anything is off by a mm or if someone is not in the room you won't achieve this and (as Marco Lopez said) end up with high frequency residual at best, or mutilated dialogue and no cancellation at all at worst.

Maybe, with very clever and tedious spectral editing (Izotope RX) you might be able to make the music less obvious and then maybe you can put louder music over it in an attempt to mask it. This might be worth a try, but depending on how much headroom there is between the original music and the dialogue, adding something louder on top will probably make everything worse.

  • Been meaning to have a look at Izotope RX for a while now so might be a good time to do that. Sounds like a clever idea with the masking but I do fear the dialogue headroom isn't enough and with the budget the filmmaker has, the tediousness somehow doesn't seem worth it. If I get a spare minute I'll give it a go on a 30 second section. Thanks for your response...
    – Phil Lee
    Feb 1, 2012 at 10:00
  • you could try using RX to remove speech from the source record and, then, try with phase cancellation. if won't work, you'll try with a long and tedious spectral editing under RX. Feb 4, 2012 at 16:59

I realize this is an old thread, but iZotope RX3 does a great job of this if the music is in stereo. What you can do is use the Center Channel Extractor in RX3, and it will use phase cancellation to isolate the mono dialogue from the stereo music. I work in a trailer house where we frequently need to clean up bites from movies with bad stems.


To me, this is close to impossible.

But I would try to things anyway, just to find out and learn.

1) Go back to the location and try to setup the EXACT !! position: stereo ergo speaker, mic, volume, direction , you name it. Then try phase cancellation in post. 2) try to mimic the room (speakerphone, frequency, room´s reverb) then phase cancellation in post. 3) Maybe a lil bit of Izotope´s RX advanced. But no real experience here with that.

If this doesn´t work and would be the best idea anyway As Shawn stated :

Reshoot and let the filmmaker learn out of it.

Good luck!

  • Hi Larry... as above no chance of the re-shoot. Some interesting ideas there but its a no go... cheers...
    – Phil Lee
    Jan 31, 2012 at 15:12
  • 3
    Also, i believe this kind of phase cancellation doesn't work. Anything less than an impossibly exact level of exactness will only boost the music. Jan 31, 2012 at 18:00
  • There is absolutely no way you could recreate the recording to do phase cancellation. If the temperature of the room is off by half a degree (from the original recording) it would be different enough to not work.
    – user22688
    Oct 18, 2017 at 18:55

I would say impossible because, the music it's a complex signal( frequency, level of each frequency, harmonics) that blends with another complex signal the dialogue in a complex space (reverberation, early reflections, some phasing perhaps or comb filtering).... So, my wild guess is that if you ever manage to recreate the same conditions and you start phasing out the music, there's gonna be some residuals there before you start affecting the dialogue. That's one scenario. The second one: How much do you charge for this one?

I agree with Shaun that the reshoot is THE option, and perhaps the cheapest.

  • Hi Marco... the re-shoot is a no-go as I mentioned to Shaun above. It is going to be pretty well impossible to do. As much as it pains me I think I'll turn down the job as I think the outcomes will destroy the original feeling of the interview which in itself it totally fine to use. If I get the time to have a go i will and keep you all in the loop... cheers Phil
    – Phil Lee
    Jan 31, 2012 at 15:16

This has happen to me a few times too and my approach to his is similar as yours. If your music is broadband you will need to both edit and spectral removal. When possible I use to remove the music between dialog lines and fill it with clean room tone. The rest will need spectral repair. For this I feel comfortable with iZotope RX.


This solution isn't entirely "honest", but you could suggest the filmaker get a good voice actor and just reconstruct the interview with new audio. Yeah, I know, it's kinda lame and would get harder if the interviewee appears more than once but perhaps it'll fit the directors needs.

Also this is my first post.. hello everyone.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.