software to reduce reverb?

[Baluardo]

Hi all,

in the past i already unsuccessfully posted this question in other forums. But somewhat i consider this context the most reliable, so i wanna try again.

- Is it possible to eliminate recorded audio reverb in post production?

we have software able to reduce noise and do fancy kinds of filterings... nobody could develop anything to reduce reverb?
Looks like books and search results tend to get rid of the topic very quickly like "it's just not possible" but it sounds to me like they dont just bother wasting time with it since there are workarounds (lavs, blankets on the walls)

But in some limited budget situations (shooting in small rooms, close to the walls, when you can't really shield walls with blanket nor get the boom closer nor can you afford lavs..) you just can't help getting a lot of reverb in the recorded conversations...
I guess such a software would be really successful if existed, and also would save a lot of complicated setups when getting audio on location...

Thanks for the possible answers-
Andrea

 

[discs of tron]

never heard of any such thing even claiming to exist. the thing is, there are all sorts of ways of filtering out sound in certain frequency ranges. but if the frequency you're trying to reduce happens to be a frequency that the "good" stuff uses (like the spoken voice,) then you can't do it without making the voice sound like crap.

the fact is, this whole notion of "saving complicated setups" is a very unhealthy way of thinking. "fix it in post" and all that. you know the maxims- garbage in, garbage out. you can't polish a turd. etc, etc. you wouldn't author a post called "is there any software to make ugly people pretty?" i mean, there probably are a few after effects plugins that'll help with that, but that's not my point. start taking the recording of the sound more seriously. recording it right is what will actually save the most time, not non-existant plugins.

 

[Baluardo]

well i do take sound recording seriously and i always try to do the most to get the best results in the first place. i'm aware the sound is the biggest pain the fix in post.

but it would be nice to know all the possibilities which are available. so that if i have 20min left of my shooting day for the last scene i might decide to shoot it 'wild and fast' and then fix it in post rather than renting equipment and reschedule cast and crew for another day.
Or this could save the money for renting / buying lavs when they would be used for just a couple of takes in a reverberating context which is hard to mask. I'm talking about zero budget here.

Noise filters provide a fix for poor audio recording, and i think all of us used it sometimes. Well, in the same way it would be nice to know just what's on the market about reverberation, that's it.

obviously i'm aware that you can't filter reverb cutting off frequencies, but same is for noise. and noise filters exist. they have some processing which goes beyond frequency cut.

Thx for your contribution anyway
Andrea

 

[Rosestar]

There is noise reduction software out there.

check this:

http://www.google.com/search?as_q=&...s_occt=any&as_dt=i&as_sitesearch=&safe=images


The problem is that reverb is complicated and is different for each location and even for each mic placement at that location. Simple, one button solutions do not exist, TMK. Look at the search results above and good luck.

 

[Matt Grunau]

You could try a noise gate. Since most sound is louder than the following revervb, a noise gate could help.

 

[thisiswells]

*maybe* a selective frequency phase cancellation might work.. if such a thing exists.

Generally, the results of any drastic post changes aren't exactly what you would have wanted. That's why they coined the phrase, "garbage in, garbage out" which isn't particularly encouraging for low budget situations, but "them's are the rules." There is a reason big productions look/sound/feel the way they do. They're using the best of all available resources. Just do the same and you'll be confident in knowing you do all you could with what was available, always for better, never for worse.

 

[Rosestar]

You could try a noise gate. Since most sound is louder than the following revervb, a noise gate could help.

I'm pretty sure that's what my sound guy used on some stuff we that had some reverb problems. He did a good job, but you could still tell that it was processed.

 

[wabbit]

There is no software or solution for reverb in post...period. It gets mixed up with the primary dialog on all frequencies making it impossible to remove. Many have tried to create a phase offset idea for post but the randomness of the echo makes it "impossible". I have read that some company (don't remember who right now) is trying to create some software but fundamentally it is highly unlikely to ever work and if it does the results will probably be uneven at best. On top of that the software is probably going to be VERY high end...costing $5000 or more just for that plug-in or whatever form it takes.

One thing that can help is if you have very late reverb you can cut it out where it appears between words but it will still be present and muddying up the actual dialog. In some rare cases this process will at least slightly improve your track.

Sanken just introduced a new lav, the COS22. It is a double mic (shaped like a T with mic on either end). It is suppose to be able to greatly reduce reverb by using phase offset while recording the audio (one end of the T is closer to the mouth, the other further away). You mix the two mics on set until you reduce the reverb and background sound as much as possible. Haven't actually seen (heard) one work but if it does what it promises this could be a revolution in location sound.

What everybody has said is dead on accurate. The reason we use blankets and avoid live rooms is precisely because it is impossible (as of today) to eliminate reverb in post. Just a cold hard fact.

Best

 

[Rosestar]

There is no software or solution for reverb in post...period. It gets mixed up with the primary dialog on all frequencies making it impossible to remove. Many have tried to create a phase offset idea for post but the randomness of the echo makes it "impossible". I have read that some company (don't remember who right now) is trying to create some software but fundamentally it is highly unlikely to ever work and if it does the results will probably be uneven at best. On top of that the software is probably going to be VERY high end...costing $5000 or more just for that plug-in or whatever form it takes.

One thing that can help is if you have very late reverb you can cut it out where it appears between words but it will still be present and muddying up the actual dialog. In some rare cases this process will at least slightly improve your track.

Sanken just introduced a new lav, the COS22. It is a double mic (shaped like a T with mic on either end). It is suppose to be able to greatly reduce reverb by using phase offset while recording the audio (one end of the T is closer to the mouth, the other further away). You mix the two mics on set until you reduce the reverb and background sound as much as possible. Haven't actually seen (heard) one work but if it does what it promises this could be a revolution in location sound.

What everybody has said is dead on accurate. The reason we use blankets and avoid live rooms is precisely because it is impossible (as of today) to eliminate reverb in post. Just a cold hard fact.

Best

Hey, were can we get some info about these mics... I'm bidding on a project where I know I am going to be dealing with this.

 

[wabbit]

They were introduced at NAB and I am not sure of their shipping status. I do remember the list price was somewhere around $600.

Here's a PDF on them.

http://www.plus24.net/dealers/pdf/sanken/COS-22_NEW_Med.pdf

If Dave Fisk is listening, maybe he knows more about these...how well they work, when they are available for purchase, etc.

 

[Baluardo]

Wabbit: i dont have a whole lot of experience with lavs... but i didn't think echo was a big deal when using lavs, generally...

Andrea

 

[robroysyd]

It IS possible to reduce echo in a recording. This is a very old trick and I'm amazed that no one here knew it could be done.
It does take patience but here's how it's done.
Make two copies of the track, placed exactly one above the other. Now invert one. The mixed result should be nothing if you've got it right so far.
Now start applying compression to only one of the tracks. You'll need to carefully adjust all the compression parameter so be very patient, pay attention to the threshold, you seem to need to get that just above the level of the echo. Adjustments to attack and decay times will have an impact as will compression ratio.
This trick was used even using two R2R decks, that would have been really tedious!
How does it work?
Well because the echo is at a lower level than what you want and as it's below the compression threshold the echo on both tracks in unaffected by the compression and hence continues to be cancelled. The compression does however affect what you wish to keep, as the compressor changes the wanted signal on one track it ceases to be cancelled out and bingo.
This works best of course if the wanted signal has some dynamic separation from the echo and I sure wouldn't recommend it on music but for dialogue I've obtained dramatic improvements.

 

[Baluardo]

Robro: this is actually very interesting procedure and it makes total sense to me, at least in theory. i will most definitely give it a try! thanks!

Andrea

 

[robroysyd]

Hope it works for you. I've done it a couple of times using Vegas which makes it very eay to try changing the paramaters for the compressor etc whilst you monitor the results. Don't expect to get perfection but it can make the difference between a ruined shoot and something with at least intelligible audio.

 

[wabbit]

Lavs pick up reverb as well as any other mic. The thing that makes them somewhat better..sometimes, is the inverse square law applies when you are able to get the lav mic closer to the speaker and further away from the source of reverb. Thereby making the primary sound source louder then the reverb then if you had used a mic (boom mic) further away.

Robrosyd, I have read about that approach to removing reverb somewhere (DV.com?) but never tried it out. From what you describe it would remove a huge chunk of the primary vocals as well as the echo (depending on extremity of the reverb).

Do you have some samples of what you have achieved with this method? I would love to hear what can be done.

best

 

[Baluardo]

I quickly tried Robrosyd's method last night on a very reverberating recording. i couldn't really improve it in any way, but i admit i was moving around the controls of the compressor in a pretty random way..
i got the impression that was easy to get rid of the echo when the main voice was 'off', but i couldnt really improve the quality of the voice itself.

but i'll try it more carefully soon.
Andrea