It seems most of my footage, which we recorded on Sony HDCAM tapes, is poor. There is distortion in a lot of the dialog, which sounds very tinny or has annoying echoes. (There's also a lot of static and background noise, but I know Audition can fix that.) I've been told that there's no way to fix distortion if the audio has been recorded poorly to begin with. Just thought I'd ask the experts here if you agree or there actually is a way to fix it. (If not with Audition, any other possibilities?)
TheodoreG, Mar 24, 2009 10:53 AM
You can make the sound slightly less unpleasant to listen to with some EQ and even noise reduction I guess. Static and background noise can in some cases be removed pretty well especially if it is continuous; i.e. a soft fan noise is possible, but varying static and random noises is in practice very difficult or impossible. It's a budget / time thing too; Audition have some amazing surgical tools for you if you have the time, skill and persistence, but even they can only do that much. 'Distortion' is actually missing data, and you can't bring forward something that simply isn't there.
>'Distortion' is actually missing data, and you can't bring forward something that simply isn't there.
Well that's true, but not by any means the only definition. Distortion can also be caused by too
much data, and not all of it is removable by any means. Basically, it's mixed in with what's wanted - and like cakes, you can't exactly 'unbake' them after the event.
The real issue is that in a system that by absolute necessity has to be automated (far too much variable data to be analysed by hand), the automated system has no real means of identifying what's wanted and what isn't - with a few relatively minor exceptions. In order to remove a component in this scenario, it has to be capable of isolation in an electrical sense - so for instance, with a stereo signal, it's not impossible to establish what portion of it is at any given position in the stereo field, and remove just what's at that position. Which means that something like a vocal, which is generally in a single fixed place, can be operated on with a reasonable degree of success (although anything else in that position will also be affected). But the same principle, when applied to something like reverb, simply won't work - because reverb is invariably decorrelated, and occurs all over the soundfield. And because it can't be isolated, it can't be either treated or removed.
There are various things that
may in the future be possible, but they are going to require massive amounts of processing, because at the very least multiple tracking FFT systems will be required.
EQ in the highs will round off the harshness. Too much and the voices will sound dull. A method that SteveG has suggested in several instances is to use a (parametric) EQ with a high peak and quite narrow Q and manually sweep through the audio as it plays to find where there's the
most noise, i.e. perhaps the opposite of what you'd expect. Then flip the curve so that the noise is removed. You will need to do this in separate and multiple passes - keep copies of the file!
The Noise Reduction is great at removing continuous noise. You take a representative noise sample (from where there's no voice - just a little snippet) and the filter will then attempt to remove this specter from your signal. Here again it can be valuable to work the opposite way; you have a button 'keep only noise' and with this you can listen for a 'clean' noise, i.e. not modulated by voice. When you flip and remove the noise you can be more safe that it affects the voice less. Amazing tool - but you will probably need to put some work into it regardless.
For static and sudden bursts switch to Spectral view of the waveform, and here you can use the Healing tool (looks like a band-aid) and paint over the bright dots that most likely are noise bursts. Again an amazing feature, but the end result will depend on your persistence with the mouse. Have others listen to your work now and then. The realistic goal you can reach here is better intellgbelity and perhaps a 'non offending' sound. What sounds bad for you (after hours of editing and focus on the problems) might be perfectly acceptable for others (client?).
(Sorry for any poor spelling etc. in this post. English is not my first language, and now it's even laaate here...)
I'm brand-new at Audition. Am processing 78s for a radio show. Now working on a 1939 Kay Kyser Brunswick 78. Judging by your previous posts I'm on the right track cleaning up background noise, clicks and pops. But I'm coming up to deal with some physical distortion of the groove caused by previous repeated playing with a heavy stylus. Harry Babbit's wave form is distorted. Maybe nothing but a cleaner copy of this record will help. Any initial ideas?
Mike WDUQ Pgh PA
Some video cameras will artificially create high frequencies from
the actual lower frequencies in a poor way so that in the spectral
graph, you can see a mirror effect, as if the lower frequencies
were folded again above some line such as 4KHz. If you see this,
you can resample the audio (to 8K in the above example) to get rid of
the bogus frequencies.
If the audio is intended to be mono but has two tracks, distortion
will often appear in both tracks but at a slightly different time.
In this case, the Center Channel Extractor can help. This can
work with vinyl mistracking, too.
When distortion occurs as a result of too high a recording level
(for instance, video cameras / cell phones overloading from heavy
bass at a concert), data is lost. Clip Restoration is supposed to
help here, but often is not too useful. You can try shaving down the
offending frequencies, or substituting from the other better track in
the case of stereo.
Sometimes, the Click/Pop Eliminator can help.
You can use Noise Reduction to isolate offending noise, then use
the Dynamics Processor to get it under control, then replace it.
If you can detect the periods of an echo, you might be able to reduce
it. If Audition allowed taking a noise sample from a script (but it
doesn't), you might to able to reduce reverb.
Clip Restoration. In EV, go to Effects>Restoration>Clip Restoration (process). I've used this a couple of times when levels have gotten out of control and clip. It's not perfect, and certainly using proper recording techniques is much MUCH better, but this can take a little bit of the edge off. Again, this isn't a magic wand program and it's not going to cure really henious distortion. Record your levels properly, that's the only magic wand.
> Am processing 78s for a radio show. Now working on a 1939 Kay Kyser Brunswick 78. Judging by your previous posts I'm on the right track cleaning up background noise, clicks and pops. But I'm coming up to deal with some physical distortion of the groove caused by previous repeated playing with a heavy stylus. Harry Babbit's wave form is distorted. Maybe nothing but a cleaner copy of this record will help. Any initial ideas?
Hey, I listen to your show, "RS&H". Regarding worn 78s, sometime around two or three years ago there was a very informative program on "Hot Jazz Saturday Night" (WAMU Washington, D.C.) that featured a couple of guests who had recently completed a big project of restoring 78s for a new CD release. On the show they described their process and played some comparisons of several clips of past releases of the music that had been on LP and various CD reissues. Their new material was, of course, the best.
Aside from digital restoration matters, their most significant technique was in very careful and possibly unusual selection of styluses. They aimed to find a stylus for each record that did not rest in the worn portion of the groove. In their experiments, they had found that the groove of a difficult record was generally worn at a particular depth, rather than completely worn at all depths. The idea is that one or more previous owners of a given record used a given stylus that had worn the groove at a particular depth over the course of decades. Therefore, to restore it, try many styluses and determine which avoids that worn part of the groove.
Back in the digital world, one thing to keep in mind is you may find it desirable to combine digitally some information from one record+stylus combination with information from another record+stylus combination. In other words, if you get your hands on more than one record of a given title, both may have useful things to offer; and both may have different things to offer with different styluses. In Audition, combining things over time (basic splicing) and at various frequencies is all feasible.