Creation of missing voice frequencies?

A

armadillo

Well-known member
I am wondering if a software plugin or program exists that can do the following...

It sound like something that can be done in theory, but I am not an expert in sound, audio, harmonics etc...

As we know, if there is brutal noise in an original recording, if you 'noise filter' to the extreme, it can all be removed but takes a lot of the actual voice with it when pushed to that level.

Is there a program out there that can take this scrubbed voice audio with missing frequencies and fundamentals etc...and generate an approximation of the missing frequencies (based on what audio it has) to "fill out" the voice sample?
 
I am genuinely curious as to whether such software exists.

Believe me, if I could ADR or re-shoot, I would. I inherited some footage and ADR / re-shoot aren't possible.
 
It also seems like an interesting theoretical question.

Again, I'm a novice with sound theory...but could one do this manually to an extent?

That is, take the existing audio track with the missing frequencies, then overlay it with itself but on the 2nd version, manipulate the frequencies toward the range that was obliterated with the noise filter?
 
My guess is that it can't be done. Take it to the extreme. A text to voice converter. Ever heard one of those sound good enough for production dialog? Sure, you can "approximate" but that's all it will be. If a computer or software could do it, why would we location sound recordists invest in the large inventory of expensive microphones in order to capture the immensely complex sound which is the human voice?
 
Seriously speaking now. I do not know of any one particular product that will automate the process.

I just finished cleaning up a 21minute short where the DP and director completely regarded the audio as a behind back burner priority didn't even have a sound guy at all , and there were lots of different problems, from mic across the room sound to no wind protection on mic, and everywhere in between. noise reduction plug-ins can help but I found the best results in using graphic dynamic plug ins and EQ sometimes treating every word or phrase differently. Also utilizing volume automation to level out things properly to get scenes match. without hearing a sample its very hard to know what to recommend, one thing for sure dont overuse the noise reduction, its good but it cant do miracles.

Use noise reduction plug ins up until you can start hearing noticeable quality drop in primary audio (you can only polish a tu&d o much), then find other methods to reduce the noise in between the words , such as volume automation with a layer of room tone and musical sound scapes to cover up the gaps in audio in between words or phrases. noise reduction is NOT your only choice.

Well the DP's and Directors lack of caring about audio cost them 40hrs of clean up in post. hind sight is 20/20, in retrospect they now could have afforded 4 sound guys on the shoot.

I guess everyone has to learn their own way. ok Ill jump off the soapbox (sorry)
 
Cool, appreciate the tips.

Yeah, not expecting miracles...

But using your extreme example with the voice synthesizer from text... The current situation does have the advantage of beginning with half a real voice or whatever...

I am getting the picture that this doesn't exist...but at least from the viewpoint of my limited knowledge...it does sound worth exploring...
 
I am thinking that if I attempt to overlay the audio track with the frequencies pushed into the missing ranges... It won't be perfect but MIGHT provide a 10-20% improvement or something...

Anyway, since audio filtering seems to be an art of "as much as you can until it sounds bad" - this may possibly create more wiggle room and allow another 20% or something of the noise to be filtered out if some of the missing range is at least partly replaced...
 
you could duplicate the track , add a pitch down filter + low pass and mix both track back together.
 
Last edited:
What do you mean by graphic dynamic plug-in?

recordingwhiz said:
Seriously speaking now. I do not know of any one particular product that will automate the process.

I just finished cleaning up a 21minute short where the DP and director completely regarded the audio as a behind back burner priority didn't even have a sound guy at all , and there were lots of different problems, from mic across the room sound to no wind protection on mic, and everywhere in between. noise reduction plug-ins can help but I found the best results in using graphic dynamic plug ins and EQ sometimes treating every word or phrase differently. Also utilizing volume automation to level out things properly to get scenes match. without hearing a sample its very hard to know what to recommend, one thing for sure dont overuse the noise reduction, its good but it cant do miracles.

Use noise reduction plug ins up until you can start hearing noticeable quality drop in primary audio (you can only polish a tu&d o much), then find other methods to reduce the noise in between the words , such as volume automation with a layer of room tone and musical sound scapes to cover up the gaps in audio in between words or phrases. noise reduction is NOT your only choice.

Well the DP's and Directors lack of caring about audio cost them 40hrs of clean up in post. hind sight is 20/20, in retrospect they now could have afforded 4 sound guys on the shoot.

I guess everyone has to learn their own way. ok Ill jump off the soapbox (sorry)
Click to expand...
 
Izotope is a great tool, or Soundsoap.

Graphic Dynamics is a common plug in, I use the one provided in Vegas but they are available from other makers. its basicaly a dynamic processor that is controled by and X,Y line. if you understand how it works you can treat it like a noise gate, downward expander, compressor and limiter, depending on how the curve is drawn. only thing is you have to know how it works to use properly.
 
Well AU Pitch certainly is not the greatest.

Did some programmer play a practical joke when naming one of the parameters "Anal Window Shape" ?
 
armadillo said:
As we know, if there is brutal noise in an original recording, if you 'noise filter' to the extreme, it can all be removed but takes a lot of the actual voice with it when pushed to that level.

Is there a program out there that can take this scrubbed voice audio with missing frequencies and fundamentals etc...and generate an approximation of the missing frequencies (based on what audio it has) to "fill out" the voice sample?
Click to expand...

No. When people 'clean up' incredibility noised audio, its because the plane went down, and they are trying to determine the pilot's last gasps from the black box recorder. The audio is processed to allow the investigators to attempt to reconstruct the crew's last efforts to avoid the crash.

Removing the 'noise'... or even if it is a clear as a bell sound, it is an 'unwanted signal', will affect the wanted signal, and so you are left with some form of compromise, and if there is a lot of noise, there will be a lot of compromise.
 
no plugin can do what you desire without altering the original. Whatever you do will always be a compromise and in many cases there is simply nothing left but to re-record the audio!
Filtering with the steepest filters will give you artifacts and if the noise is broadband there is nothing you can do.
 
You must log in or register to reply here.
Back
Top