Audio levels all over the place, can you help?
 

Ok folks, I was shooting some footage of a guy training some techs on how to use a controller interface for an HVAC system and we were shooting in the basement of a building with massive compressors and airhandlers and the noise was ear splitting. I had a wireless lav on the guy and everything is going along fine and then the guys starts bobbing his head left and right. at the extremes of either end his voice is barely audible, as his head approaches middle it gets clearer and then in the middle it gets hot and then back again. i have been trying to manually clean this up boosting the low peaks and lowering the high peaks but i am wondering if there is something better I could be using to fix this...some sort of peak normalization software, plug-in, or tool. Any ideas?

I am editing in Vegas Pro 9a and using Adobe Audition 1.5 for clean-up... Thanks for any help here.

Brian,

Not that would be any better than watching the waveforms and making manual adjustments.

Recording in noisy environments is a b*tch.

Were you using a cardioid patterned lav?

Regards,

Ty Ford

Trying to fix this sort of thing is very individual, dependent upon the specific track. I can tell you a few things I have used which may or may not work for you.

You have three problems:

1. Ambient noise
2. Level variations
3. EQ variations caused by the person going on and off mic.

Copy the track twice so you have three identical tracks. One is to be the 'mostly OK' track. One is the 'way off mic with no highs' track. One is the 'in between the two' track, or maybe the 'too much bass' track. The idea is to process these seperately, then recombine them.

So use your level automation to remove all the parts on one track that don't match the 'no highs' problem. On the second track, remove all the parts that are not fine, on the third, remove all the parts that are not in between ot too much lows. Now if you played all three tracks at once, you would get what sounded like the original back.

So now you can process each track separately.

First, try to remove as much noise as possible from each of the tracks. There are threads on this board that mention various noise reduction plugins and techniques for this. Also, downward expanders help.

Next re-eq the tracks to sound as good as possible - so you can now, for example, boost the highs on the track that is too dull sounding to match the 'normal' track, since only the dull portions are on that track. Probably the lows will have to be reduced on that track as well. etc.

Then you can use a slow compressor on each separate track to fix the level problems.

Finally you re-combine the three, creating once again a single track.

Another approach, if you have a multiband limiter available, is to boost the highs on the track so that the normally 'too dull' parts sound ok, which will make the previously ok parts sound too bright. Then send that through the multiband, with the threshold set so it will knock the too bright portions back down to nomal while leaving the boosted dull parts alone.

Hand leveling, as Ty mentioned, works best, and will help hold down unwanted artifacts from compressors and other processing if you do that first. But often the changes happen too fast for hand leveling alone to work.

Also, doing fast cross fades between the differently treated tracks can help fix the intermediate cases and will sound more natural than hard on/off switches between tracks.

Sometimes a mild compression on the whole combined track will help things re-blend back together.

A de-esser might knock down some of the impulses from air hammers.

It can be a lot of work...

-MD

Huh? I'm talking about doing this in post where you can see the waveforms on the timeline.

Regards,

Ty Ford

I thought you were talking about fixing the level problems by manually adjusting levels (as opposed to using a compressor) - now I am not sure what you meant. ;-) (the normalize function? Hand editing the waveform? I'm confused...)

...Oh wait, now I see, of course it cannot be 'too fast to do by hand' in that situation. Old way of thinking on my part. Nevertheless I would still use a compressor to handle the fast stuff rather than make a zillion little hand changes.

-MD

In FCP , Pro Tools and most audio programs I've seen, there's a line that indicates the volume level of a track. You can drop points on it and raise or lower the audio quite precisely.

I just did this within FCP for a 5 minute trailer with a LOT of sound files; dialog, music and SFX, all w/in FCP.

http://idisk.mac.com/tyreeford-Public/HotFlashMovie.mp4

Regards,

Ty Ford

I know that, it is just my age showing. I have many more years of pushing faders around than mice, it is just wired in. I still tend to write automation mostly by hitting record rather than using the mouse to move the line around, etc. It's faster for me.

-Mike

Age is no excuse! :) I think I have you beat in that dept.

Regards,

Ty

I'm 59. Senility begins early in my family. ;-)

Sounds great, looks good - but what is this? Is there an actual movie? You have a producer credit?

-Mike

The noise reduction portion worked out pretty good, I was able to clean out most of the undesirables. Thankfully, the subject paused a lot so i had plenty of opportunity to extract noise profiles. The big issue is the back and forth head motion. It did not help matters that he was wearing a T-shirt and I did not have my vampire clip with me so I had to clip the lav right at the neckline. I angled the mic away from the noisy equipment but that means when he turns his head to the right I get a severe drop. On waveform, I am looking at a peak variance that ranges from around -3dB to about -21dB at the lower peaks and since it is a back and forth motion the in betweens are really hard to control. Unfortunately, the ambient noise was so loud that as i monitored with headphones I could barely tell what was going on with his mic until I got it captured (I was about 15 ft closer to the compressors and air handlers than he was.) So, if I am reading the suggestions correctly there is no silver bullet that will analyze the track and equalize the peaks, manual is the way to go...

I was asked to produce this previsualization (previs) to help the writer sell either the script or to get enough funding to shoot the entire script.

I picked a director, LD and helped with casting. I did location audio, most of the post production and wrote the lyrics for the theme. It took up most of my August in a very fun and challenging way.

Regards,

Ty Ford

What did you shoot on? Do I detect your voice doing the VO?

Sounds like a really fun month!

This should be in a separate thread in the Techniques for Independent Production section. It looks like a pretty darn good way (and example of a way) to sell a script.

Is it common now to do something this extensive? I normally think of 'previsualization' as something considerably less polished.

-Mike

Correct. Normalization or a leveler can help even out peak levels, a multiband EQ can help make the highs and lows more consistent, but there is still going to be a lot of manual work no matter what.

-MD