Removing boxy room tone?

[Dave Bourke]

Hi all,

OK, please don't bite -- I'm essentially a composer still trying to figure out all this arcane engineering stuff. Maybe if I had another lifetime or two...

Anyway, I have an accidental recording of my young nephew telling his four year-old little sister a fairytale. Part of it is perfect as an intro for a track I'm almost finished. I'll never be able to get that innocence in a \"performance\" for the mic so I can't record him again.

Problem is, it was recorded in a long narrow hallway about 4 feet wide and the ambience is boxy and horrible. How would I go about getting rid of it? I know a little bit but I wouldn't even know where to start in removing a room tone. I'm using a UAD-1 Studio PAK (along with Plate 140, Precision EQ, and Precision Limiter) and a Powercore Firewire (along with Assimilator and VoiceModeler).

The sound I'm after is either an outdoor ambience, or a nice warm, open room.

Any takers?

Kind regards.

 

[ed1966]

As far as UA plugs go the Cambridge EQ would be the best bet to go after the room resonanace. If you're a PC user you can download Voxengo's SPAN frequency analyzer to help you visually locate the pesky frequencies.

Probably the best tool out there for this application is Har-Bal (currently PC only).

If the audio is reasonably short, I'd be glad to do that for you for free. You can track down my email thru audiorecordingandservices.com

 

[brian]

You probably need to do multiple cuts with a parametric EQ. Cambridge or P-EQ will work well for this. Experiment around the 350-500Hz range. If you find an area where the cut is effective, chances are you can look at other multiples of this frequency for resonances from the hallway (ie: cut at 400hz, and also maybe 800hz). Depending on the recording, a HP filter may be appropriate. You can try adding some roomlike ambience with the EMT plate set for a very short and slighly dark verb.

 

[Dave Bourke]

ed1966 and brian,

Thank you both for your very helpful advice.

I tried just applying a verb room preset over it but it just sounds crap. So the answer is to cut away the crappy frequencies first and THEN to apply the reverb. For some reason I was thinking backwards, guys. So thanks for setting me straight

I'm on the Mac, ed1966. Hope I'm not the only Mac guy who checks out the PC side 'cos I didn't post this in the Mac forum. Anyway, I can't use SPAN or Har-Bal, but I do have Elemental Audio Systems' Inspector. I'll take a look with it.

Many thanks for your kind offer to do this free for me, ed1966. The piece is about 50-60 seconds long, I think (I still haven't decided how much of it to use). But I really have to start doing this kind of stuff for myself and learn by trial and error. If I screw it up, I may very well come back to you for further advice if that's OK?

Now that I have a clear handle on how this is done, I'm gonna give it a go. After all, I can't break anything...or can I?

Thanks again, guys. I appreciate your help.

Kind regards.

 

[akisd28]

You could try EqSat for cutting the unwanted frequencies. I've found it's the best EQ for cuts in the TC/UA arsenal (IMHO etc.).

 

[Dave Bourke]

Thanks, akisd28. I'll give that one a shot as well. What the heck, I'll try them all -- that's why I bought them, I suppose

I have to admit something weird: i've only recently got all these amazingly cool and powerful tools and I'm kinda scared to start really using them in case I screw up. Dumb, huh?

Kind regards.

 

[akisd28]

It's almost certain you'll screw up at first, but hey, how do you expect to learn otherwise? :wink:

 

[boody]

you can also try an expander plugin like floorfish (digitalfishphones.com).
You use it to keep the voice and kill the boxy tail, put a nice reverb on it and correct the freq of the voice with the new eq or cambridge before warming it up with a nice compressor and a pultec to make it deep and silky.

 

[cAPSLOCK]

While you are getting 'floorfish' get 'dominion' from the same place. It can also help you with this.

cAPS

 

[toader2]

I have an idea... it's probably not very practical - in fact, I'm not sure how you could do it... maybe somebody out there knows if it's even possible though.

Using the current crop of convolution reverbs, they can use an impulse to simulate a room...

I wonder if they could use the same impulse to remove the sound of the room...

This way, you could put the mic in the exact same position you had it in when you recorded your nephew speaking. Then you could put a speaker in the approximate position of your nephew's mouth. You could play an impulse sweep through the speaker, and record it with the mic. This way, you know you have an impulse of the sound of the reverb in the hall...

Maybe using this recorded impulse, there is a way to do a \"negative convolution\" to remove the sound of the room. Maybe there is a plugin that can take the reciprocal of a file. This way, you could take the reciprocal of the noise print, and load it into any convolution reverb, and then process your original recording with it.

Maybe this is ridiculous... maybe insane or impossible... maybe not though.

Todd

 

[Cass Anawaty]

That's similiar to using phase cancellation to remove frequencies. Just cut the voice out as much as possible, double the track and invert it. Not really practical, but I have heard folks do some pretty amazing restorative work along these lines......
CAA

 

[toader2]

That's similiar to using phase cancellation to remove frequencies. Just cut the voice out as much as possible, double the track and invert it. Not really practical, but I have heard folks do some pretty amazing restorative work along these lines......
CAA

That's kind of the idea I was thinking of... but it always "sort of" works... never a really perfect solution. Maybe it would be a significant improvement though... that's what got me thinking about new ways to approach the issue.

The more I think about it though, maybe the reverb would be too decorrelated from the original signal to be able to do any kind of "deconvolution". Oh well, just at thought.

Todd

 

[toader2]

I just did a little test...

I took a dry vocal, and processed it with a room sound using an impulse and the SIR reverb. I then inverted the phase on the impulse file. After that, I processed the dry vocal again, but this time with the phase reversed impulse file.

When mixing the two resulting files together, ALL of the reverb cancelled out - 100%, leaving the vocal perfectly intact and dry once again!

You may not get perfect correlation (from the impulse file, and the natural ambience on the recording of the nephew).

However, if you try the method I mentioned above, I bet you are able to cancel most of it! If you have the patience and capability, I would love to know if it really works.

Todd

 

[Mark Edmonds]

I was wondering about this method as well. You have proved it in the digital domain but in a real life scenario where the chances are, the exact recording setup wasn't noted? I'd say fergedaboutit Even if you mange to get a close match on the impulse, aren't you going to start introducing all sorts of odd phase and timing errors? Wont it end up sounding like a demented flanger or something? I'm guessing here BTW so please correct if necessary!

Not knocking the idea, just saying I would be surprised if it can be got to work.

Mark

 

[stratcat]

Hi Dave,

I'd be surprised if the cambridge can't scoop out the boxiness.

probably between 400z-1K ?

good luck

SC

http://www.soundclick.com/bands/3/peter ... ymusic.htm

 

[Dave Bourke]

Wow! I never expected to see all this assistance offered to a raw beginner. But between all the generous people on this forum and the Powercore forum I've now got enough tips to keep me experimenting for days (and learning, too, which is what I'm most in need of). Thank you all so much. In particular:

boody: thank you for the \"floorfish\" suggestion. The Fish Fillets package looks pretty tasty. Oh, and thanks for the recipe

cAPSLOCK: Unfortunately, I can't use Dominion since it's MacOS 9 or Windows only. And I'm on Mac OS X. However, I read up on what it does and I think I might be able to use PSP Nitro for this (unless I totally misunderstood what I read). So thanks for the suggestion.

toader2: brilliant suggestion on using a convolution reverb. Two problems: I don't own one (and I won't buy Altiverb). And my sister (the child's mother) was the one standing up against the wall in the hallway hiding the mic behind her back. So that can never be replicated. But what a great idea!

Cass: thank you for your help here and on the \"other\" forum. I find phase cancellation baffling as a concept (I spent too much time dreaming out the window in maths and science classes). But I'll do some research (I never realised before just how much skill and hard work engineers have to put into a job).

Mark: I think your observations are spot on. As I said, I won't ever be able to replicate the circumstances of this \"accidental\" recording, so I think it rules out the convolution reverb idea, good as it is.

stratcat: thanks for your advice on using the Cambridge for this. I've also been hearing very good reports on Motu's new MasterWorks EQ which now comes stock with DP. And I've also got TC's stuff to play with as well: Assimilator, Dynamic EQ, EQSat, MasterX3, and VoiceModeler.

If I can't do it with all the UA and TC tools, then maybe I should just stick to tinkling the ivories and leave this stuff to people who know what they're doing, and why

I am indebted to you all. Thanks a million for taking time to pass on the benefit of your experience. Unfortunately (or fortunately?) paying work reared its delightfully ugly head this morning, so I won't be able to report back on how I got on with this until next week.

Meanwhile, thank you all once again. You'e a grand bunch altogether

Kindest regards.

 

[Animus]

You might try somehting else which I think might work like a charm. Sample the room in silence with the same mic you used. Take the resultant file and run it though one of those noiseprint cancellers such as Waves (get the demo). I have had great results doin this and no need for all the convolution stuff.

hope that helps



Just thought of something else even more simple..... If there is a short period of silence in the actual voice recording you can just use that as a reference for one of those denoisers.