I have a recording that was done with a camera mic, it was a retirement party so there were speeches, but as there was no dedicated mic for the speaker, the camera mic picked up not only the speaker, sounding distant and a little distorted with some echoes and reverb, but also some conversations from people close to the camera. They should have obviously used another mic, but that is the past.

What effects/compressors/normalizers, etc. should I use to try to bring out the speaker's voice and minimise the unwanted conversations to make it the best it can be? I've already used the "noise removal" effect in audacity.

  • Your question (and a few similar ones) have sparked a larger discussion on the topic here; sound.stackexchange.com/questions/40005/… I think an "expander", as suggested by frcake, is your best bet here, but unfortunately, you might find this task nearly impossible due to the points in the post I linked.
    – user9881
    Commented Oct 10, 2016 at 19:26
  • Thanks.. the answer given by Marc W really reminded my of this: ted.com/talks/… which is amazing on its own, I couldn't comment there, I don't have enough rep. you could if you wish :) but I think my problem is much less acute.. I don't need to separate sources of a mix, just to lower the background noise and dry the distant speaker.. i'll try an expander again, Thank you :)
    – Ray
    Commented Oct 10, 2016 at 19:49

2 Answers 2


1st of all, if the voice coming from the speaker has some characteristics picked up by the fact that's miced and going through amps & speaker, maybe you could EQ those so they can pop a bit and create a contrast between the other sounds/people talking w/e..

If the "right next to the camera" talkings obscure the speech, sadly there's little you can do. The main thing that you should aim for is frequency separation to start with, if that isnt doable, dynamic separation is not doable either. (in your situation)

Now if all the above fails, and you really want this to happen , I would do it by hand, edit out any part that obscures the speech, it can be painful but could save some stuff.

But look for the frequency "footprint" first, see if you can boost that against the others.

EDIT after OP's comments.

To explain my answer further :

When you have a sound that you want to manipulate, for instance make it dryer (which is not such a simple task on its own) and separate it, the classic way is to boost it's unique frequency response and cut everything else.

As a simple example I could easily use a kick drum. To separate a kick drum from a hihat, or a snare, first of all you would dump the frequencies that you don't care about, and then use other stuff to separate more, like gates or compressors..

But if you try to separate a kick drum from a bass track, it's nearly impossible. Though you will be able to only play sound when the kick drum hits (with a simple gate), you wont be able to take out the bass sound from the sample/waveform. That's because they have very similar frequencies.

Now in the kick drum example we use a transient, a kick sound. With continuous sounds (such as talking) the job is harder by far.

The main body of the voice is somewhat close in frequency response so you end up with a midrange thing.

So even if the expander worked, when the expander would open you'd hear the talking on the main voice. Cause it just works like a gate - it opens when it listens to the sound and the sound by itself is not separated!

But the fact that you say you are sure it can get separated intrigues me, so to further advise I would like you to upload a sample of the problem. It's the only way we can have a more detailed discussion :)

  • Thanks @frcake, the main thing is that the main speaker is furhter away, but is heard. there is background noise and a litter chatter and sometimes also the "right next to the camera" talking which only a little lower in volume than the main speaker. I have to make the main speaker stand out. I've tried EQing. it didn't get me the best results. now can you think of anything else that can be done? Many Thanks!!
    – Ray
    Commented Oct 5, 2016 at 6:54
  • Well you could use an expander or if you can compress the speech i'd duplicate it and try to phase cancel it.. Compressing will get the vocal out of the signal leaving the rest to be nulled frim phase cancelation.
    – frcake
    Commented Oct 5, 2016 at 7:02
  • After some clarification by OP, it appears that this is a single mic/ one-track recording, but they're concerned with removing background noise, so an expander might do the trick.
    – user9881
    Commented Oct 10, 2016 at 19:19
  • @frcake Thanks a lot! :) I'll try some things and let you know, Thanks so much for your proposal :)
    – Ray
    Commented Oct 12, 2016 at 22:35

At first glance it seems like you need an "expander", but upon further reading it seems like you're trying to separate different sound sources from within a single recorded track.

Unfortunately, you are probably not going to have much luck separating such similar sounds as different people speaking despite any volume difference.

  • Thanks for your answer, I don't mind being able to isolating all speech and removing all background noise/echo etc' which approach should I try? (already trying to figure out how to "phase cancel" it like @frcake has offered. many thanks to both of you!
    – Ray
    Commented Oct 6, 2016 at 16:23

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.