Audio - problem with "normalize"
 

Two questions about audio --

First, occasionally when I use the "normalize" switch on a poorly recorded audio track that needs to be amplified, it flatlines -- the track becomes completely inaudible and the waveform is just a straight line. I've tried rebuilding peaks, but that doesn't seem to help. Does anyone know what might be the problem?

Second, and I'm a novice at this, so forgive this question, but what do you guys do when you have a bunch of video clips and associate audio clips, and the volume levels of the voices are different from one to the next? How do you keep it consistent from one clip to the next?

Up to now, I've been normalizing the audio tracks, so that the loudest part is set to the maximum level, and this works well except for the occasional hiccup like my previous question where normalize doesn't work at all. Also, this method doesn't work when there is a loud non-voice noise (like a cough or someone dropping a book or something), because that louder noise becomes the maximum level for the normalization, and everything else is set from there. The result is that the track turns out too quiet compared to the other tracks.

Any advice or pointing me to a good basic text would be appreciated.

Hi Brian,

Welcome to DV-INFO. What I have done when you have noises like cough, loud sound, that's killing your normalize function is this. Locate the high transient waveform(usually at the beginning of such noise) and select just that portion. Use an amplify effect set to -3db which will halve the volume. When you get that peak to a level that fits the rest of the audio, then the normalize will work better. It can be tedious to find all the high peaks first, but the method will work.

good luck,

=gb=

Your first issue is a known problem that sometimes happens. Not sure WHY it happens but it sometimes does.

As for adjusting volumes, I just use the various volume controls. I may boost the track volume some and then use the volume envelope to set the actual volume needed. I can boost it up or down in various areas as needed that way. I don't worry about normalizing.

If I wanted to normalize (and really fine tune the volumes as precisely as possible) then I would use Sound Forge, adjust as needed, and bring the adjusted version back into Vegas.

It's a very tedious process, but, I just finished reducing 7 hours of DV to a 5 minute promo. Clips from 3 different cams, the sound guy had problems with his audio recorder so I had to use audio from each camera. It was a nightmare to balance the audio. I did it by using the volume envelope. Had to go thru the video and do it manually one clip at a time.

Your first issue is a known problem that sometimes happens. Not sure WHY it happens but it sometimes does.


The SFK reference file may have been interrupted in its creation (ie doing something else on the machine as vegas builds peaks) another reason this occurs is when u begin to access said clip on the timeline during the peak building phase.
Way to fix this is to delete the clip and reimport it to the timeline. That works for me, but i hardly get this problem now.

one thing about normalisation is that it scans the WHOLE clip, so peaked areas will affest the lower ends. Easiest way to deal with this is to split the clip and then crossfade it, and run the normalisation again on teh isolated quiet areas. this ill rebuild th escan and normalise properly

So, on a DVD project with several motion menus and multiple clips, all with different audio levels, one approach to bring the levels into the realm of similarity would be to place them all in Vegas 5 one their own tracks, then go through and using a combination of track levels and volume envelopes, match them all, letting my ears be the guide. This could be done by comparing two tracks, setting one completely Left channel, one completely Right channel, and playing through, iterating through the lot. Then select each video/audio pair and Render To New Track, thereby creating new versions with the matching audio levels?

What other options are there? Anything automated? Are there any numerical or "absolute" ways to compare the levels?

Sorry for the ignorance. I just played what i thought was the final draft of the DVD and discovered the large differences in audio levels, particularly between motion menu tracks and the clips they pointed to.

BTW I haven't overlooked any simple audio tools in DVDA2 for adjusting levels have I? I couldn't find any.

thanks for shedding some light,
Scott

Well... you could add all the clips in Vegas and normalize each one. That will get the loudest point in each clip to the same level.

Peak levels are meaningless when it comes to determining how "loud" your final product is. The human ear doesn't determine loudness by the peak level, it determines it by the average (or RMS) level. The "normalize" function in Vegas is useless because it only makes adjustments based on peak levels and that's not they way we hear things.

If you're watching a movie on TV and it has some relatively quiet dialog and then a commercial comes on, the commercial sounds louder because it has been heavily compressed in order to raise its average level (and get your attention.) The fact of the matter is that the movie and the commercial probably had the same peak level, it just that the commercial has a much higher average level.

Audio compression and limiting is an art form and it takes a lot of experience to do it "correctly." There are no hard and fast rules to determine the appropriate average level, you'll just have to play it by ear. But like I said, peak levels are virtually meaningless (well, as long as they don't exceed 0db.)

You could use the Normalize function in Sound Forge. Render out the audiio from the entire timeline and pull it into Sound Forge, then mark each different clip or section that needs to be level matched. Use the Normalize function in Sound Forge and set it to normalize using "Average RMS power" and normalize to -20db and select "If clipping occurs apply Dymanic Compression." -20db is a good starting point, you will have to play with this value for your project. Once you have determined the appropriate RMS level, use that same value for everything and the loudness of will be matched throughout your entire project.

John

EXCELLENT reply, John. This is the first reply that makes sense. Thanx.

OK, I get it, Agerage RMS power sounds like the term indicating, in laymen's terms, the 'loudness' I hear, at least relatively speaking. it's exactly like the difference between television programming and commercials. I heard the volume of the DVD menu background, then when i moved to one of the clips, I could hardly hear it's audio. So the RMS power of the clip is significantly lower.

Actually hear are the numbers based on running the
Statistic Tool in Forge:
Menu Background RMS power: L(-16.86dB) R(-19.43dB)
Clip RMS power : L(-21.48dB) R(-21.50dB)

So I Normalized RMS to -16 dB, applying dynamic compression for clipping, on both clips.

now the Statistics tool says:
Menu Background RMS power: L(-14.22) R(-16.78)
Clip RMS power: L(-19.15) R(-19.17)

The two clips are still significantly different in terms of volume. Any thoughts? I promised delivery tomorrow, which is no one's problem but mine. But the ink still has to dry on the DVD's once I burn them. Ha.

Scott

I've noticed the menus always seem louder than the content.