Need suggestions from pros on how to process low-level audio

[Asterra]

I thought I more or less knew what I was doing when it comes to audio, but I have to admit I am stumped.

I used a portable recorder (Zoom H2) to record fireworks a few years ago. The nature of the audio is that it is essentially as raw as it gets. Explosions sound muted despite hitting between 0 and 1 dB, and dialogue hovers at a near inaudible ~50 dB. What I am trying to do is process the recording to make the audio sound more like something recorded in a standard fashion (with a camcorder, for example), so that it is a good match for the other audio I'll be using in the same project.

I'm using Adobe Audition, so the obvious choice is the dynamics processor. I've made use of it on several occasions, and while I have never been fully comfortable with the attack/decay nature of the process (and the occasionally blunt artifacts this produces), I still figured I'd be able to use it to turn my 24-bit recordings into something that sounds better-rounded, with explosions and dialogue equally listenable.

The biggest hurdle I am running into is the apparent disconnect between the dynamics curve I set up and the result Audition kicks out. Let's take, for example, a curve which contains the following results:

-96dB -> -96dB
-50dB -> -15dB
-20dB -> -10dB
0dB -> -10dB

Now, to my mind, this curve should enforce a -10dB limit on the output. But Audition seems to feel that it can ignore my curve and clip the heck out of the output. No input or output gain. It clips away.

Maybe I had better just provide a sample of the audio I'm working with and see if anyone can suggest a process series which will turn it into something more camcorder-like, hopefully without obvious attack/decay artifacts. The clip will include some dialogue, some particularly loud explosions, and a sequence with "Saturn" missiles which I have found especially challenging. All are from the same recording. I have not found any series of processes which renders the lot down to something entirely listenable.

Good luck to anyone who decides to help a poor soul out! And many thanks!

port_clip1_24bit.wav

[Cornucopia]

Ok,
Not quite sure what you're needing to mix in with this, but here's my take on it...

You've got clipping, so you need to start by doing clip restoration. Did that, lowered by ~4db.

Then, I added dynamics processing, kind of like what you were talking about, but I set the gain so:

* Note that I set the attack & release times to be very short (1ms A, 50ms R, maybe shorter on the level sensing) and linked the 2 channels.

Then, I added a mild bass boost for oomph on the bursts and a little improvement for the vocals (parametric EQ - beefy snare preset to start)

Then, I did a normalize to -0.333dB.

Is this what you're looking for?

Scott

edit: also used Audition.

[Asterra]

Thanks for taking the time to help. Making progress now.

I notice you've run into some of the issues I've encountered and have had to carefully skirt. For example, at 00;22;14, the second explosion in a sequence of four is basically missing its entire attack. I've needed to tweak things a lot to avoid such occurrences.

I'll give an example of the kind of sound I'm after. This was achieved with the following graph and settings (after a -4dB clip restoration which I forgot to mention I'd been doing):



The resulting audio is very heavily clipped (fair warning!) but it is pretty close to the sound you'd get from a camcorder recording the same events.

port_clip1_24bit_edit1.wav

Now, obviously the problem here is the extremely heavy clipping. I know of a few ways to approach the problem.

1: Restore clipping again. Well, this would reduce the overall amplitude, and in the example above, it's close to where I want it (the saturn missile sequence could use some further reduction in RMS and the four loud explosions need to have their overall loudness increased a bit (but not their attack!)).
2: Do another dynamics process. Well, see, I was sort of hoping that my 0dB -> -10dB specification in the first process would have taken the bite out of all of those millisecond-long attacks in the first place. It didn't. (I'm not actually looking to limit the amplitude to -10dB; it's just my exaggerated attempt at a solution to the issues I'm experiencing.) Besides, I have found that doing two dynamics processes on the same audio tends to surface considerable audible artifacting.
3: Hard limit? Unfortunately this process has shortcomings. A minimum decay time of 40ms means that each one of those tiny, clipped attacks forces the audio to be briefly subdued before returning to normal levels in an audibly gradual fashion.

All three of these possibilities result in something that sounds worse than the heavily clipped result I've provided above. Said result is sadly the best thing I've managed to get out of Audition so far. What I hope can be found is a way to get the same levels out of the dynamics processing but also kick those huge blips (the 1ms start of the explosions) down to manageable levels without introducing complications like the one I pointed out in the .wav you uploaded, or audible attack/decay manifestations.

The reason why I feel like this shouldn't be as difficult as it's turning out to be is because if I were to simply play this 24-bit recording over a loudspeaker and record its output with a camcorder, I actually would probably get precisely what I'm after, without this bizarre artifacting. I feel like this hoop-jumping ought not to be necessary. ;p