Getting constant audio levels across multiple events (help)

newbie123 wrote on 6/25/2003, 7:41 AM
i am in the process of putting together my first masterpiece, aka p.o.c. when i look back on it in two years, but for now i am happy and learning.

i have inserted multiple audio events to the timeline but each audio event has a diferent volume. this is what i have done so far, please tell me if there is an easier / better / smarter way of doing this.

i have added audio envelopes (i think that's what their called) to the audio timeline. this is the blue line that you can move up and down with points you insert. What i have done so far is carefully tried to insert the points wherever the volume drops below or above the -7 to -10 range to try and get a constant audio level throughout the entire event. I have done this with three audio events so far. I have left them on the same track, would it be better to put them on seperate tracks?

i tried the normalization but couldn't get it to normalize the multiple events as if they were one, and ended up with one really loud event and the others not as loud.

i have played with the plugins and to date the envelopes is the only thing i have found.

does anyone have any suggestions or ideas?

by the way, this video is actually a wedding video for friends of mine that i filmed almost two years ago, but only recently got a computer stable enough to edit on.

thanks.

dgg in ottawa

Comments

Chienworks wrote on 6/25/2003, 8:02 AM
The volume envelope is probably the most straight forward and simplest way to achieve what you want. The problem to watch out for is that if you increase the volume too much and it goes over 0dB you'll get distortion. Remember that louder events can be reduced in volume just as easily as raising quieter ones.

As far as normalizing goes, you don't want to normalize the entire audio track as a single event. Normalizing affects the entire selection uniformly, so if you normalize everything together, the louder parts are still exactly as much louder than the quite parts as they were before. Normalizing each section independantly brings each up to the loudest it can be, which may achieve what you're after. The problem you would have with normalizing is that if there are any very loud peaks in an otherwise quiet section then normalizing only raises the level up to get the loudest peak to maximum. There are several ways around this. One is to simply split the event just on either side of the peak and then normalize all the smaller sections individually. Another method is to apply compression before normalizing in order to reduce the peaks. These methods are more complex than using only a volume envelope, but they may give better results if used properly.
mcgeedo wrote on 6/25/2003, 8:06 AM
Do you know how to set up loop markers, to play a loop over and over? If not, read the manual and learn how to do this first.

Then set up a loop on one of your audio events. Play while watching the audio level meters. These meters are by the master audio level controls. Watch the average level, indicated by the meters, and also the audio peaks, indicated by the numbers just above the meters. Adjust the audio envelope for the looped event to get the peaks to read 0 dB or just under.

Then go to the next event, loop it and adjust the envelope for that event. When you get done, all of your events should be near the same volume (level). Then you can fine-tune the envelopes so that they actually sound similar.

More advanced: Normalize your clips before you put them on the timeline. Study up on how to use the compressor plugin. The compressor will help you get similar average volumes rather than adjusting based on the peaks, which isn't as accurate.

I hope this helps.

-Don
newbie123 wrote on 6/25/2003, 8:37 AM
thanks for the feedback. from what you are telling me i seem to be on the right track, just seems to take a long time to get it. but then again most things do.

i have done the looping trick and played with the points and i think i have gotten a faily decent and somewhat constant audio level.

now to pick the brain a little more. should i be trying to get the master volume level to approach the 0db level as much as possible without going over. ie spikes hitting the 0db level, or is trying for a -7 to -10 range and having the peaks hit the -5 and -4 a safer way to go so as not to get any distortion.

my theory was that they, my friends, could always just turn the sound up initially if they had to.

thanks

dgg in ottawa
mcgeedo wrote on 6/25/2003, 9:05 AM
That's a bit more complicated. First, there is quite a bit of what is called "headroom" above the 0 dB point. So your peaks can easily go to +3 or greater and you'll never hear any distortion. For speech, I usually set the levels so that 0 dB is just touched (and sometimes just exceeded). Then I use the compressor plugin to lower the peaks 3 dB or so. This makes the average a little higher, and usually improves the understandability of the speech.

Now you need to determine what media you're going to put your final print on. If it is DVD, then you probably want to set the overall level down some. A DVD has a very high dynamic range, i.e. can handle a large difference between the loudest and softest sounds. Hollywood DVDs are usually set so that speech is fairly low, to leave room for the volume to increase for car chases, explosions, etc. So to make your DVD about the same level as a Hollywood DVD, you need to set your average level to something quite a bit below 0 dB. I use around -12 dB. The idea here is: Put a Hollywood DVD in your player. Set the volume to a comfortable level. Take it out and put in your DVD. The average level should be similar.

For VHS, on the HiFi tracks, you do something similar. For the LoFi tracks on a VHS tape, you can leave the level fairly high and the recorder will compress it for you.

Good luck,
-Don
newbie123 wrote on 6/25/2003, 9:39 AM
thanks a lot for the tips.
mcgeedo wrote on 6/25/2003, 9:45 AM
No problem. I'm not one of the "old pros," but I have faced the same problems you described, and found those simple answers.
John_Cline wrote on 6/25/2003, 3:45 PM
Don,

It is never a good idea to go over 0db. Unlike analog audio, there is absolutely no headroom above 0db in digital audio. While you may not be able to hear the clipped peaks by raising the audio to +3db or more, rest assured that they are indeed clipping. If you were to open the finished audio in an audio editor and zoomed in on the peaks, you would see them flattened. This type of clipping is extremely rich in harmonics and increases the total harmonic distortion considerably.

If you really need to "punch up" an audio track, it is always better to use a limiter plugin like Sonic Foundry's WaveHammer or Wave's L2 UltraMaximizer. In fact, you could use the track compressor in Vegas using a high compression ratio, like 10:1 or even 20:1. You'll probably also need to set the attack and release times a bit faster as well so it just clamps the peaks and immediately gets out of the way.

John
thrillcat wrote on 6/25/2003, 4:15 PM
Am I the only one on here using the Wave Hammer plug-in? I love that thing. It has a compressor and a Volume Maximizer all in the same plug-in. Kind of a combo Compressor/Normalizer. It compresses the clips to get them in the same ballpark, then maximizes the overall volume to a level you choose (personally, I use -3db).

I would suggest getting the clips close (if they're exremely different) with the volume envelope, then run the Wave Hammer plugin on the track.

mcgeedo wrote on 6/26/2003, 8:36 AM
Hiyah, John,
I agree wholeheartedly that one should never go into digital clipping, for exactly the reasons you list. While I often use 0 dB as a reference point to make clip-to-clip levels consistent, I always master down to something a lot less, like -12 or -14, to render.

But I have a question: Is 0 dB indeed the absolute peak, i.e. representing "all ones" in the digital word representing a sample? Since the peak-reading meter can go above 0 dB, it would seem that 0 dB represents something less than the absolute max binary value.

This is probably an important enough question to justify making a test. I'll let you know what I find.

Thanks, John. I appreciate your thoughts.
mcgeedo wrote on 6/26/2003, 9:31 AM
Well, I made a test, and I find a curious thing. In a 16-bit wave file (according to spec), the maximum positive voltage is represented by 7fff (in HEX), 0 is 0000 (HEX) and the maximum negative voltage is ffff (HEX). This is the absolute maximum voltage (level) that can be represented. So a nice clean sine wave (tone) should start at 0000, ramp up to a maximum of 7fff, back to 0000, down to ffff and then back up to 0000.

I made a sine wave file (with another program) and loaded it into VV. (I use VV3). I set it to a level (per the meters) of 0 dB and rendered it out to a 44,100, 16-bit wave file. Looking into the file with a binary editor, I see that the maximum positive voltage is represented by 3fff, i.e. just half of the maximum possible value, i.e. 6 dB of headroom.

Sounds reasonable so far, but now the funny part: I set the level to +3 dB and rendered it out. Looking at that file, I see that indeed it is digitally clipped, and pretty severely so. But the fascinating thing is that the maximum voltage, in the middle of the clipped peak, is STILL 3fff! VV3 is indicating that this file is at +3 dB, and yes, it sounds terrible, but VV3 doesn't seem to make use of the 6 dB of possible headroom.

Do you suppose that Sonic uses the extra 6 dB internally, but intentionally renders clipping at 0 dB? Do you suppose that the meters indicate (at least above 0 dB) what the signal WOULD be, if not clipped at 3fff?

Very interesting...
mikkie wrote on 6/26/2003, 9:32 AM
FWIW, as John wrote, there really is no head room. Software metering can be misleading, & I don't think that when the levels read over 0db it should be read that way to say I've got all this room still available.

Sound Forge allows you to normalize based on rms levels, which may help a little bit for the individual clips, but I think the biggest challenge would be to try and get and get the transitions from clip to clip to sound natural re: dialog volume levels and such. You may or may not be able to do that strictly according to db levels, so make sure to rely on your ears before you render. Also don't forget that when you raise levels, you're raising the volume of any noise that might have been recorded as well.

Compressors/limiters can help a lot, but I'd suggest reading up on the topic a bit to make sure you understand the process so you wind up with what you intended. There have also been a bunch of audio tips here in the forum on massaging the dialog in your audio tracks, so try a search.

If it helps, personally I've found that setting a max db level between -2 & -3 gets you pretty close to TV broadcast settings in the US providing you don't have a ton of dynamic range - ie: if your audio sounds like a TV broadcast rather then varies between whispers and booms as on a DVD.

PAW wrote on 6/29/2003, 1:27 PM
Interesting post.

The audio side of things is new to me.

Anyone else want to chip in with their cookbook of audio levels for different media. This post has made me realise a few things in relation to audio levels - I have probably been pumping it too high.

What do people use for voice, narration, background music, effects etc for DVD, Broadcast

What should I be aiming at so that I fit in with the rest of the production world.

Cheers, PAW
24Peter wrote on 6/29/2003, 10:17 PM
Where can we get the Wave Hammer plugin?
Chienworks wrote on 6/29/2003, 10:56 PM
24Peter, if you've got the full version of SoundForge installed then you've already got Wave Hammer. I don't know if Vegas includes it without SoundForge though because i don't have only Vegas installed anywhere.

If you do have it installed, open up the Plug-In Chooser for either an audio event or track and it should be near the end of the list.
PAW wrote on 6/30/2003, 2:55 AM

You need SoundForge to get Wavehammer

PAW