How to isolate certain voices in video?


I have several home vids I took with my phone, in which there are multiple people talking as well as music in the background.

Is there any way I can isolate certain voices? If it helps, the main conversation I want is from the focal point of the vid.

Angela Dean

Posted 2016-03-30T03:39:07.387

Reputation: 1



There is no way to do what you want. Human voice, no matter what voice timbre people have, is pretty much in the same frequency range.

That said, you can at least filter out some of background noise by equalizing out frequencies that do not occur in human voice. There is a good article on voice frequency on Wikipedia:

So to filter out frequencies from an audio recording you typically use an equalizer filter which allow you to enhance or reduce the occurance of certain frequencies in a recording.

This is a good article on equalization, especially pay attention to the chapter "filter types":

Hans Meiser

Posted 2016-03-30T03:39:07.387

Reputation: 183


There will be no good technique for this, in your case.

They best you could do is to use and EQ to see if you could emphasize the vocals and deemphasize the background.

The common technique for isolating vocals from a song is phase inverting the background track to cancel it out(this won't work for you).


Posted 2016-03-30T03:39:07.387

Reputation: 674

I don't know why this answer was downvoted. – Hans Meiser – 2016-04-03T11:19:04.377

Because the bulk of this answer doesn't have anything to do with the question. And the part that makes sense doesn't explain much. He'd make a good politician! – Marc W – 2016-04-03T14:46:05.810

I give the 100% accurate answer right at the start. And then some related background info about the best way to do such a thing. Proper answer imo. – Scorb – 2016-04-03T16:24:54.307

But that "background" has nothing to do with the question. As you stated; "this won't work for you". Vocal extraction via phase cancellation is a completely unrelated process, so why post a video? If you had left that part out, and explained the filtering part more, like @HansM did, I'd have up voted it. I'ts currently not a good answer imho. If you edit it, I'd be willing to reconsider my vote. – Marc W – 2016-04-03T17:20:37.077

Yes but the answer is "nothing will work for you." So the "answer" plus "background" is not any worse than the answer..... – Scorb – 2016-04-03T17:36:28.223


The only thing I would add to Hans Meiser's answer is the possible use of an expander/gate.

If the target conversation is the focal point, then it should be slightly louder than the rest(at certain frequencies), allowing you to use an expander or gate to emphasize this conversation. You may find it tricky though, If you aren't an audio professional.

You would need an expander or a gate with internal(or external) sidechain capabilities (Example).
You could then find the frequency(or frequencies) in which the target conversation is dominant, and feed that refined signal to the sidechain.

Then, with some adjustments to the expander/gate's parameters, the result should be a signal that gets louder and quieter with the target conversation's voices.

In practical application, this could be tricky to maintain. It could take a bit of work.

What is an Expander/Gate?
Apple explanation of Expanders/Gates
Pro Tools Tutorial: Sidechain Techniques
Side-chaining in Cubase

Marc W

Posted 2016-03-30T03:39:07.387

Reputation: 2 042


It might be difficult to isolate a certain voice. In Audacity, you can import an audio file and split it. Then select the lower part and select effect and select invert. However, it will not isolate all the voices.

Also, you can filter the frequency but again, that will not be as precise since your audio changes its frequency over time (unless the one who spoke is monotonous).


Posted 2016-03-30T03:39:07.387

Reputation: 1

This doesn't really give an answer, as It's most likely a monaural signal. – Marc W – 2016-04-03T14:51:56.520