Removing slight echo with audtion - recorded in room with hard walls

[Guest]

Hi

I just downloaded a trial of Audition CS4 for windows.

I have already used it to remove a few clicks which are evident in this short (attached) PRE-audition sample.

Now I would like to go in and remove (if possible) or at least reduce the slight "echo" of the room itself.  You can hear the "echo" in this attached wav file.

I realize I might not be able to remove all of the "echo" but if I could remove some of it it would be nice.

Thanks

Rowby

[ryclark]

Unfortunately it has all been said further back in this thread. There is no easy way of removing echo or reverb. There is one slight chance but it would rely on the echo being very precisely constant time away from your original. You could try copy and mix pasting your full level audio delayed by the exact time over the echoes to cancel them out. I have a feeling that this probably won't work because the delay time of the echo may be variable throughout the piece.

[AskMrScience]

I'm curious about the interval.  If it's as long as 5 - 7 seconds, is this by chance a tape recording?  Sounds like a case of magnetic print-though.  In any case, I just did a quick web search on "echo cancellation", and there does appear to be deconvolution software out there, meant for things like Skype, that may hold promise.  I haven't used any of them, so I can't report any first hand experiences.  Within Audition, no, I can't think of anything except perhaps noise gating.  Good luck!

[Bob_Howes]

I must say that 5-7 seconds sounds long even for a phone line.  Are you sure you're not being bugged? 

In any case, the sort of echo you get on a phone is very different from the room reverb that was being talked about before and you may have a slightly better chance.

If it was me, I'd probably try using a gate, carefully set to allow the "main" voice through but closing for the echo level.  I'm not usually a fan of using a gate for most noise control but on an echoey phone it can work okay.

Failing being able to get a gate to work, I'd just make a cup of tea and sit down with the file, muting the echo manually (which is, in effect, becoming a tea powered gate with automatically variable threshold control).  Or, you might end up with a gate setting that gets most of it with only a few leftovers to attack manually.

[rmartin215]

Thanks to everybody for a quick reply. I'll try everything that I can.  Since the audio is only 13 minutes long, I will go thru it manually.

The call was from India.  I think it had something to do with that, because local calls have no echo.

Thanks again.

[Bob_Howes]

Almost certainly because it was from India and transmitted via satellite.  Assuming you are in the US, that's a minimum of 2 satellite hops to get from India to the US and two more to get back.  Each hop is about 2/3 of a second delay so the time could add up quite quickly.

[denis1]

What where you trying exactly? If you feed *only* reverb into UNVEIL, that won't work very well, as it'll try to discern between direct and reverb - so if there's no direct components the process has insufficient information. That's a testing sceanrio that just won't work.  Also, the FOCUS control reduces reverb moving it *up*, and increases reverb when moving it *down*....so you may actually have *told* the plugin to leave only the reverb, instead of removing it. Also, the assessment that "it is not a plug-in" makes me wonder whether operation may have been unclear to you in more general terms. (it IS actually a plug-in, there's just ALSO a small app that comes with it on Mac...). We're glad to help you get more out of UNVEIL, just email us!

UNVEIL does actually work very well, though we're not doing magic - it will obviously not always be possible to remove all of the reverb in a recording. Also, it does not affect a) discrete echoes or b) room induced resonances, which are quite often misunderstood to be the same thing (while all of these phenomena boil down to reflection-induced repetitions, their mathematical properties differr greatly).

Feel free to contact us with an example file that you're not getting good results with using UNVEIL, and we'll send you a plugin preset that works.

[SteveG_AudioMasters_]

denis@zynaptiq wrote:

Also, it does not affect a) discrete echoes or b) room induced resonances, which are quite often misunderstood to be the same thing (while all of these phenomena boil down to reflection-induced repetitions, their mathematical properties differ greatly).

I'm fully aware of that. Unveil doesn't do anything which I have any use for, personally - I don't get in these messes in the first place...

[Allynn Wilkinson]

Hey Denis,

I was just on the Zynaptiq page and noticed a new program, "Unfilter".  Can you tell us the difference between Unfilter and Unveil?  I've used Unveil and found it helpful but I wonder if Unfilter would be more useful in the original case mentioned here. 

It would be great to never have to worry about these probelms, bue, unfortunately, some of us have to.

Just found a great review by my guru, Larry Jordan:

http://www.larryjordan.biz/product-review-zynaptiq-unveil/

[MusicConductor]

SammyT303 wrote:

Just read this and man your an *******.  He was just trying to help.

I'm not sure who Sammy was addressing, but now I see everyone else has dogpiled on this to say how stupid it is to dredge up a 4-year-old thread to act like... a 4-year-old.

Steve, adding your opinion about religious matters as fact does not add credibility to your reliable, scientifically-based assertions.  Just sayin, friend.

[Bob_Howes]

Much as I agree with SteveG's comments on religion, I guess it might have been inappropriate to bring them into these U2U forums.

However, I'd point out that it was joe12south who brought Jesus into things in the first place.

[Joe12south]

Actually, I was making a sideways reference to a Sarah Silverman joke. 🙂

Regardless, it is possible to improve recordings, at least somewhat, with reverb baked-in. There are several tools and techniques to salvage otherwise unusable clips, and it's good that people know that.

[Bob_Howes]

Much as I'd love a product that could eliminate early reflections, despite what you say I've yet to hear one that did what it said on the box.  In my experience, usually it sounds worse after processing, particularly in terms of intelligibility, rather than better.

Maybe someday there will be an ultra-clever software solution but none of the simple stuff mentioned (like EQ) does anything other than mess up your sound.  Like SteveG, I hate inexperienced people being mislead by well-meaning people with bad advice.

[Joe12south]

The product I mentioned does an admirable job in certain situations. The trick of extracting a center channel from a stereo mic does a good job with vocal recordings...there are options for "rescuing" clips that otherwise might be unusable. Perfect? No. Better? Certainly. The OP wasn't looking for perfection...heck, the recording wasn't that great in a lot of other respects...he just needed to salvage it. We can argue academic "impossibility" or we can try to offer real-world help.

[SteveG_AudioMasters_]

MusicConductor wrote:

Steve, adding your opinion about religious matters as fact does not add credibility to your reliable, scientifically-based assertions.  Just sayin, friend.

The whole point is that it wasn't an opinion. What I said was only based on facts subsequently established by scholars. And I didn't mention the guy in the first place...

[MusicConductor]

The whole point is that it wasn't an opinion. What I said was only based on facts subsequently established by scholars. And I didn't mention the guy in the first place...

I know you didn't, and I also know you've seen more than your share of the wretched affects of the abuse of religion, a damnable and very un-Christian problem (regardless of what "label" the religiosity has).  However, "scholars" can be depended on to "prove" most anything, and there is massively much evidence to the contrary for me and many others to agree with yours.  We may have to agree to disagree on this one...

[SteveG_AudioMasters_]

Starts OT, but gets back on track:

MusicConductor wrote:

However, "scholars" can be depended on to "prove" most anything, and there is massively much evidence to the contrary for me and many others to agree with yours.  We may have to agree to disagree on this one...

Firstly it's not mine, it's theirs, although I know a present-day one. FWIW, she would be massively more inclined to your point of view than mine, I have to say. And, the physical evidence of what appears to be scholarly abuse approximately a couple of thousand years ago has disturbed her somewhat...

Where's the parallel? We appear to have software where the claims about it can be 'interpreted', and indeed have been. As with all of these things, there are some Laws of Physics involved, and you ignore them at your peril. All I am doing is trying to be objective - as scientists (and scholars...) really should strive to be.

And as for 'impossible' - which was the basic problem this time around - well, some things pretty much will be until some of the Laws of Physics get changed, I'm afraid. The problem with reverberation is that of reconstructing a wavefront at the point it was launched, and removing everything else that happens to it between the source and where it's being received. One very major problem with this is that although the radiated waveform is essentially from a point source (therefore spherical) in the room, by the time it gets to the microphone the angle of the sphere and therefore the shape of the wavefront has altered considerably. The physics behind this is complex; suffice it to say that even with really posh equipment, you can't easily calculate what even the launch wavefront looked like with any accuracy at all - and that's without even considering the impact that every reflecting surface in the room is adding to it.

Whatever you get from 'reverb removal' is going to be fakery for the forseeable future, I'm afraid.

[Guest]

For what it's worth, here's a suggestion made on another forum

"You can try using Effect -> Reverb -> Convolution reverb... to remove echo: Create a new track using the convolution reverb effect and play it against the old track, with phase inverted (search help for Invert a waveform about that).

What settings you will use and if this will be successful or not, depends on the settings that you will enter at Convolution reverb filter and if you're lucky enough to match the echo. There is no success guarantee!"

Any thoughts about this approach?

Rowby

[AskMrScience]

Adding reverb in an attempt to remove it?  You'd be adding a new layer of reverb to the existing unwanted reverberance was well as to the signal that you're trying to isolate.  It quickly becomes a hall of mirrors - and phase shifts.

[SteveG_AudioMasters_]

rowby wrote:

There is no success guarantee!"

Any thoughts about this approach?

Indeed - I can guarantee no success!

If you want to read a sensible explanation of why this won't ever work, then I can refer you to the old FAQ about this that used to be on this forum, but is now only available (like all of the rest) on AudioMasters. It's here.

[Joe12south]

This plug-in does the so-called "impossible" quite well: http://www.zynaptiq.com/unveil/

It's pricey, but if it saves even one session, it can be worth it.

[SteveG_AudioMasters_]

Joe12south wrote:

This plug-in does the so-called "impossible" quite well: http://www.zynaptiq.com/unveil/

It's pricey, but if it saves even one session, it can be worth it.

Yeah, we did that one a while back in this thread... and it's not that good.

[Joe12south]

It's not magic in the way Jesus is, but it does work well in a lot of cases. I just hate when people say something is carte blanche "impossible."

[SteveG_AudioMasters_]

Joe12south wrote:

It's not magic in the way Jesus is, but it does work well in a lot of cases. I just hate when people say something is carte blanche "impossible."

Well, the acoustician in me says it is. The best you can ever manage is to disguise reverberation somewhat, by reducing the tails - perhaps... You can't get rid of the early reflection signal at all; the actual 'sound' you end up with remains coloured the way it is in the original, and that's part of the reverberation that you are absolutely stuck with - period. You might get away with doing some treatment on speech, and hiding some of the decorrelated room sound, but on music signals with clearly delineated reverberation, Unveil simply doesn't work at all. They claim that it works at least in part by using a masking process, and that shows - you get results akin to a low bit rate MP3!

Oh, and Jesus wasn't magic either. he's just been very badly reported by people who wrote about him 50-odd years after the event, and all the other reports that conflicted the four we generally get got 'hidden' by the local Ministry of Truth some time after the event.

[Kost7]

he's just been very badly reported by people who wrote about him 50-odd years after the event

50 years after event? It's still way better than other non-religious ancient manuscripts which no one doubts somehow today.

[SteveG_AudioMasters_]

Kost7 wrote:

50 years after event? It's still way better than other non-religious ancient manuscripts which no one doubts somehow today.

Can you provide an example, please?