Free software to remove "sound pattern" from recording?


I am trying to remove undesirable sounds from a recording:

I have a video conference recording that includes sounds such as "Has joined the conference" and "has left the conference" that I am trying to get rid of. The difficulty is that these sounds sometimes appear while someone is speaking.

There are part of the recording when these sounds occur and no one is speaking so I could be able to extract these sounds.

Are you aware of any free software able that would be able to match and remove a specific audio track within a larger audio recording?

All I can find online is how to remove "background noise" and "clicking noise" but nothing that would match a specified sound.

Thank you!


Posted 2016-12-08T14:27:04.213

Reputation: 113

1I don't even know of any expensive software that could do it. This is a razor-blade task; one at a time, by hand. – Tetsujin – 2016-12-08T20:10:28.413

I am also not aware of any solution to this problem, so rather than talking about a solution, I'd like to discuss the problem. Hopefully some programming genius might some day be inspired to take this on and create a solution. An additional problem regarding removing a sound is the volume and harmonics of other sounds in the area to be modified. Simply subtracting out a waveform is not going to work where that waveform is overlapping other waveforms. A spectrum view of how subtraction works: [Subtracting one waveform from another](https://i.stack.imgur.c

– Dale Mahalko – 2016-12-18T10:53:20.567

I'm sorry, but this is utter nonsense. – Mark – 2016-12-18T10:58:26.243

Which part do you disagree with, how simple subtracting of a waveform works, or the idea of a 2D visual spectrum editor? – Dale Mahalko – 2016-12-18T11:05:24.363

First and foremost, this is not the place for a discussion on whether this is a potential answer or not. It's either an answer to the problem or it isn't. IMHO your answer uses lots of long words without really addressing the issue in a manner that communicates that you understand either the concepts or the problem. – Mark – 2016-12-18T11:09:45.200


  • Not all answers on stack exchange are a direct response to the question. Some provide additional detail related to a topic not directly asked in the question. 2. "Long words" are necessary because the alternative is a huge pile of drawings or equations to explain in far greater detail. 3. My general assumption is that if you've used 2D or 3D spectrum analyzers then you know what audio looks like using them, and you generally understand how this would visually work. If you haven't then you won't. Best to just ignore this and move along.
  • < – Dale Mahalko – 2016-12-18T11:29:16.383

    I'm sorry, after having read this multiple times I can't see anything redeeming whatsoever in your response. Again it's intended to be an answer to a problem. My opinion is that it it isn't. Even the first line of your response gives this away and your understanding of how spectral editors work is flawed. – Mark – 2016-12-18T11:56:53.510

    Dale - this really isn't how the process works at all. Most of what you have written is misguided and incorrect, as Mark pointed out. – Rory Alsop – 2016-12-25T22:21:52.807

    "The remaining portion "falls down" with the bottom side becoming straight and the top and bottom curves combining to form a new shape" What? Talking about shapes really confuses things, to the point that this is very inaccurate. I recommend reading this:

    – None – 2017-08-22T02:06:54.517



    This might be possible with a some kind of electrical or mechanical noise etc. but with human voice (even synthesized voice) the frequency range is simply too wide to remove without also removing most of the fidelity of the recording in question.

    That being said, Reaper DAW (free to trial) has a built-in plugin 'ReaFIR' which will allow you to capture sort of "audio signature" and then invert the phase during playback to filter it out. Feel free to give it a shot, but I don't expect the results to be very easy-listening.

    Some tutorials exist for this, like this one (Full disclosure: I have not reviewd that video for quality).


    Posted 2016-12-08T14:27:04.213


    1Thank you for your answer! I have given it a try and you are right; as the noise that I am trying to remove is actually a voice of someone talking the range of frequencies is just too wide and too close to the other frequencies used by the actual speakers. When I subtract these frequencies from my overall recording the voices of the other speakers become really robotic and "weird". Anyway, thanks for introducing me to Reaper! It could definitely become handy when trying to remove other types of sounds! – Yannickv – 2016-12-13T11:49:49.680


    It will be a manual process of:

    1/ isolating a clean version of the sound you want to remove 2/ lining up the clean version with an instance of the sound you want to remove - sample accurate - in a new track 3/ inverting the phase of the clean version 4/ summing and bouncing the result.

    It will be very difficult to get this right, as you are most likely also dealing with compression artifacts. The result won't be high quality, but it's your best bet. There's nothing out there that will do this for you - not that I am aware of anyway.


    Posted 2016-12-08T14:27:04.213

    Reputation: 7 535

    Thank you for your answer! I have just tried that using the software recommended by DoritoStyle. The result does sound really strange and I do not think there is much that can be done other than re-record the talk of the different speakers. Thank you for the quick answer! – Yannickv – 2016-12-13T11:54:41.843