Interviews: Too live, and a goofy solution?

[Doc West]

I'm trying to de-room some interviews, and I've got a technical question
and an esthetic question.

Backstory: Recorded ~20 hours of interviews with surviving founders of a
club I'm in.For reasons too embarassing to admit, recording was to
Masterlink via a room mic and a handheld ( 1/2 of a Crown Sass-P and an
SM58). Room was a "conference room". about 20'x30', fairly live.

PROBLEM: Strikingly life-like recording includes AC, some room ring,and
the very occasional passing train. Lots of air hiss. As I explained to
interviewees "sounds like somebody bugged your living room".

SOLUTION ?: A Logic plugin called direction mixer". this can work as an
M/S decoder, but I've set it to feed Left-to-Right and blended to taste.
Result seems to minimize background noise, balance stereo better, and
still maintain stereo location.
When collapsed to mono, room noise increases, but is still better than
original recordings.

TECH question: Why is this working? I don't trust my guesses. but I
think something useful and startling is going on. This works much better
than parametric eq or expanders.

ESTHETIC question: I like the over-live recordings. They give a real
"you-are-there" vibe, which might be an extra treat 20 years from now.
But I feel compelled to make these sound more like a radio show.
since these are only for historic purposes, how far would you go towards
clean-up if it was up to you? Even in the original staate, all dialog is
clear as a bell.

PLEA: If there's a better way to do this, please tell me now. I've got
17 more hours of interviews and 30 hours of speakers to do.

CAVEAT: I oviously have never done this before, and know zero. All
suggestions welcome. In fact, I'm beggin' for 'em.
If this L-to-R thing is a good idea, maybe there's some Mac program that
will do it faster than real time?
BTW, the complete run of 80 sets is already sold, I'm on an only
somewhat flexible deadline.

Thanks, Doc W - ignorant sound guy.

[S O'Neill]

Doc West wrote:

> PROBLEM: Strikingly life-like recording includes AC, some room ring,and
> the very occasional passing train. Lots of air hiss. As I explained to
> interviewees "sounds like somebody bugged your living room".

BIAS's SoundSoap is a reasonably-priced ($99) plugin and standalone
noise remover. It's made for things like AC and it can be pretty
amazing. It works great in some situations, and in others it gets a
little phasey when you try to get too much out of it (eg, it only helps
a little with distortion - but that's not the same thing).

Anyway if you wanted to send me a piece of a file that has the noise you
want to get rid of, I could show you what it will (or not) do.

[Kurt Albershardt]

Doc West wrote:
>
> PLEA: If there's a better way to do this, please tell me now. I've got
> 17 more hours of interviews and 30 hours of speakers to do.

Get the mic(s) as close to the speakers' mouths as possible, most likely
with omnis so you don't get proximity effect or exessive breath noise.

[Doc West]

Thanks for the tips,gentlemen.
Sound Soap" looks,on paper, to be an excellent reason to go with OSX.
But I can't do that right now.
Certainly, I'll use omnis if there's a next time. But the mics were
provided, and the guy who was supposed to run things never showed. Lucky
I brought the SASS-P as a just-in-case. In any case, the recordings are
what they are and the history conference won't be repeated.

So I'm stuck with trying to make these recordings sound prettier. Ty?
Will?
The important thing was to get these people to talk. Sometimes they
interviewed each other, so my gig was only to ride gain. The SASS room
mic was a lifesaver here, but it still has that old "live stage play in
a 50's TV studio" quality. I hope to make the recordings sound
friendlier to modern ears.

I'm still intigued by this feeding L>R business. It works best when the
interviewee really spoke at the SM58 (instead of holding it at knee
level). Each interview must be post-treated seperately and uniquely.
In Logic 4.7, a bounce-to-disc has to be realtime.
With 20 hours of interviews, this is a bunch of work. Unpaid. Maybe
buying and installing OSX would be the gentler,easier way after all.

S O'Neill > wrote:
> BIAS's SoundSoap is a reasonably-priced ($99) plugin and standalone
> noise remover. It's made for things like AC and it can be pretty
> amazing. It works great in some situations, and in others it gets a
> little phasey when you try to get too much out of it (eg, it only helps
> a little with distortion - but that's not the same thing).
>
Kurt Albershardt > wrote:
> Get the mic(s) as close to the speakers' mouths as possible, most likely
> with omnis so you don't get proximity effect or exessive breath noise.

[Keith W Blackwell]

(Doc West) wrote in message >...
> I'm still intigued by this feeding L>R business.

I have no idea what you mean when you write "left-to-right".

Something I've seen, however, was that after using two cardioid
mics in a noisy environment it was possible to mix them to mono
with one side inverted (to get the difference, the "S" of "MS"),
and a great deal of the background noise disappeared, especially
the bass. Now, when two such cardioids are near-coincident,
this is a poor-man's virtual figure-8 mic. When you do this on
purpose, it can be very useful (the null plane can be directed
toward pesky noise sources). But surprisingly, this can give
useful results even if the mics are not near-coincident, though
you would expect phase anomalies to be an issue. Now, of course,
if the mics are very different in their frequency response, have
different patterns, etc., then the results will make sense
mathematically but you might not be able to guess what those
results will be intuitively. In your case you have a cardioid
and a baffled omni that might act like a cardioid for higher
frequencies, and if they were close to each other then a difference
between them might or might not be useful. If they were pointed
in opposite directions, you might get that figure-8 effect somewhat.

Pre-EQing one of the two mics to sound more like
the other before differencing them might help, and of course you
have to adjust levels to find what combination (45%-55%, e.g.)
gives you the best cancelation. If one of the two mics is already
putting out a signal that has absolute polarity reversed with
respect to the other, then you don't need to bother inverting.
Maybe that's what you've run into here. Hey, if it works, go for
it. But I really don't know what you mean.

Beyond that, try using Cool Edit Pro's "Noise Reduction" feature.
Try a good high-pass filter and cut out everything below, say,
120Hz (or higher). Try a band-limited downward expander or noise
gate to have all the low-mids and bass cut off as soon as the
talker finishes saying a word. Etc. You might still have junk
when you're done, but it might be easier to tell what people are
saying.


--
Keith W Blackwell

[willp17]

>
> PROBLEM: Strikingly life-like recording includes AC, some room ring,and
> the very occasional passing train. Lots of air hiss. As I explained to
> interviewees "sounds like somebody bugged your living room".
>
> since these are only for historic purposes, how far would you go towards
> clean-up if it was up to you? Even in the original staate, all dialog is
> clear as a bell.
>
>maybe there's some Mac program that will do it faster than real time?
>

You mention that the dialogue is "clear as a bell". How much cleaning
up does it really need? Can you do it without putting in superhuman
effort? If you do put in the time, will the improvement be
substantial? And, of course the ever popular: Is that extra bit of
improvement worth enough to the client that they'd pay for "superhuman
effort" billing? Just questions I'd be asking..

However, if you want to get rid of some background noise you could try
a Mac program called Spark XL. I've used it successfully for some
light forensic work. It has a "de-noise" function that allows you to
sample the audio that you would like to reduce and then apply it
overall. It won't help with the room reflections, but it could reduce
the AC, train noise, excessive "air", etc. Be aware that this will
only reduce, not remove, some background sounds. If you try to use it
too severely you'll get some annoying "under-water sounding"
artifacting that would be totally unwanted and distracting. If your
audio is already in aiff or a compatible file format, you can open it
in SparkXL and work on it. Otherwise, you can play it into Spark in
real time and create a file. Either way the "de-noise" processing
only takes seconds to apply to a file - depending on the speed of your
Mac and the length of the file. As always, YWWV. Check their website
(http://www.tcelectronic.com/SparkXL) for a demo that you can download
and use for a few days to see if it would work for you. It's usable
in OS9, too. Hope that helps.