Post audio levels

M

mrl72

Well-known member
Hey all

Just working on my 2.0 mix for my short film. Is it true that you typically produce different versions depending on how the film will be shown? For example, I created a mix that peaks around -12db which plays great in my dedicated media room (professionally installed surround system) yet on Vimeo etc. it's very low compared to most videos out there. So should I be creating two versions, one for Home/Festivals and a separate for online?

Cheers
 
Videos on YouTube/Vimeo, etc. seem to be into the "Loudness Wars" just like broadcast radio/TV.
I suspect that most "produced" videos are at least normalized to around 0dBFS, and many are likely compressed as well.
 
Yes films typically do "versions" for different playback environments. Theatres are quiet but homes are full of distractions so TV versions are usually more compressed. I suspect that Richard is right about YouTube and Vimeo. Not sure it's so much loudness wars as "filmmakers" with no experience processing the audio as if it were a pop tune, but...
 
Richard Crowley said:
Videos on YouTube/Vimeo, etc. seem to be into the "Loudness Wars" just like broadcast radio/TV.
I suspect that most "produced" videos are at least normalized to around 0dBFS, and many are likely compressed as well.
Click to expand...

I believe my audio guy normalizes to 0db, though I couldn't swear to it.

Without doing dedicated mixes for different venues I think most people listen to the output on a variety of platforms (from studio monitors to a laptop) and try to "split the difference". So it isn't optimized for any of them, but sounds decent on all of them.
 
Can you homies clarify something for me...

When we talk about "peaks", are talking about the loudest parts of normal dialogue that occur regularly, say every 10-30 seconds, or are we talking about the absolute loudest thing in the audio track, which could occur only once in a project, or at least very infrequently, or something else? Exactly which parts of the signal are considered peaks and should hit your peak target of -12, -10, or whatever else youve established? Always found the terminology very confusing.
 
Last edited:
I mean "regular" peaks. Very often I will go through and knock down individual slices of a waveform (sometimes just a 1/2 cycle) to 'normalize" the "outliers" so that the peaks will all be relatively uniform so that overall normalization will yield a more consistent result without resorting to compression, etc.
 
Peak normalizing to 0dBFS would most likely result in many peaks above -3.0dBFS being clipped when the file is encoded or re-encoded for YouTube, Vimeo, ect. This is common in many data compressed formats, MP3, AAC and such, depending on the degree.
 
I remember reading that on here before...because some freqs clip before others, right? But the meters wont necessarily reflect this? So would you say that to be safe, nothing should ever go over -3?
 
It's not so much that some freq clip sooner but that the D/A converter is creating a analog wave based on the digital signal. If you have two zeros in a row the wave has to go over that to connect the points. Since it can't go past zero it clips. -3db is the most over the wave would go so at -3 you won't cause distortion of the peaks.

Most Places you deliver to (professionally) have delivery specs you have to meet. So as a general rule broadcast will want your peak (highest point in the program) to be around -10, though I have also seen -6 with an average dialog peak at -12. A growing trend is to use a new measurement called a "dialog normal" level. You can only find this level by running your whole program through dedicated hardware or software At about $2K - $4K for the box/software it is a major pain for freelancers.

You may hear things like Program Peak, that would be the highest signal anywhere. I have also heard Dialog Peak, which is the loudest dialog level.

Some of these things should go away since they are based on the limits of analog TV broadcast and there isn't much of that left in the world today. In NTSC analog, the audio sits right next to the video and they can crosstalk. Really loud sound can bleed into the next channel up the dial and full white will bleed into the audio track of the same channel. There are BIG fines for bleeding into someone else's airspace so broadcasters got very strict with the audio limits. Not so much with white levels, that is the buzzing you sometimes hear when you see white subtitles and such on news shows.
 
Scott, that's pretty much what I've done. My average dialogue peak is around -12 or a bit under and some music during action scenes are pushing -3 to -6. While this all sounds good in my dedicated 7ch media room I now face the fact that it's way too low for online. So the question is do I need to remix everything for online or can I just raise the master and compress or use a limiter to keep it at decent levels?
 
When I'm mixing for Internet distribution, I'll typically start my mix as if I'm working for broadcast. At my old shop, the specs were for a hard limit of -10dB with a pretty squashed dynamic range (about 12dB), and I experienced that with several well-known cable networks. I'll give it a little more dynamic range for the 'Net, bit not much. The Waves L1 or L2 with Ultramaximizer is a great tool for this, and was pretty much our standard plug-in on the master bus. The only other difference is that my brick-wall limit is at -3dB or so for the 'Net. This is still not nearly as squashed as mastering for commercial CD releases, which is a good thing.

Dynamic range does still depend on the project, and narrative is going to require a little more than lifestyle/newsmagazine pieces, but the challenge with the Internet is that there are absolutely no standards or expectations outside of >0dB clipping. So, really, you go from there with whatever feels right to you. Everybody does it their own way, and the only ones who have any method to their madness... any basic protocol... are those of us who have done this professionally for film and television. You're dealing with end-users (web viewers) who are generally watching on laptops with crappy built-in speakers, and sometimes headphones/earbuds. Further, you're up against a lot of self-taught, script-to-post content creators (everybody's a producer these days, right?) who have little-to-no knowledge and experience when it comes to handling sound.
 
Just for our fund of questionable knowledge ...

Noiz2 said:
In NTSC analog, the audio sits right next to the video and they can crosstalk. Really loud sound can bleed into the next channel up the dial and full white will bleed into the audio track of the same channel.
Click to expand...

Specifically, in NTSC broadcast the audio and video carrier waves are adjacent in the channel. Some times very white video - which is at the very lowest RF level - would cause some audio receivers to momentarily lose carrier lock and generate some amount of "static" - the more white in the picuture the more "static". It's not an over-modulation issue as it is an under-modulation condition. Video in NTSC broadcast is inversely amplitude modulated (AM) (higher input = less modulation so video can NEVER over-modulate the carrier) but audio uses frequency modulation (FM) and it is conceivable that over-modulation from excessive audio levels could interfere with adjacent channels. As an aside, some channels' audio carriers actually were in the FM broadcast band and you could listen to them on your FM radio.

NTSC is a video specification: 525 interlaced scan lines at approximately 60 fields per second with phase-encoded color (and a bunch of other stuff!). It can be used, along with a corresponding audio signal, to modulate an RF carrier giving us traditional Over the Air TV. That same RF carrier can be modulated with anything else, e.g., ATSC , giving us OTA digital broadcast.

Probably more than you wanted to know and what this has to do with Post Audio escapes me! :lipsrseal

Live a life well-modulated!
 
Actually in the US you could quite often hear the raspy static caused by "super white" levels all too often used in "news bulletin" lower thirds.

How it relates to Audio post is that the broadcast Audio specks were set by broadcasters because of the FCC violations that could result from over modulated audio. Someone asked why -12 was a Peak and not zero. Well it was the implied question anyway.

You explained the video spec better but we are really saying the same thing.

My other point was that there is no reason for this restriction with digital broadcast.
 
Yep, I agree. Use all those bits; just leave some room for those stray high-energy ones or you'll get clipped! Come to think of it, that whole super-white audio noise sounded a bit like digital clipping, just not correlated with the audio peaks.

"Progress" often means just substituting one set of problems for another. Sigh.
 
Sort of a wyrd couple of questions...

1) At what point does the 5.1 (I just saw 7.1 on a recent release...) 'surround sound' happen?
2) Is there any way in the 'home' environment to bring out the dialog on the 'center' and subdue the remaining 'channels'?

On the disk that had mention of 7.1, there was a 'test program' that sent out signals, and displayed what speaker should be in operation.

Since I only have a 'stereo' setup, the sounds came from the appropriate left or right side, depending, with the 'center' split between them.

However, it occurred to me, if I had a real 'surround sound' system, that I could somehow attenuate the 'purely' left/righ/front/back, while keeping the dialog 'center'.

There's a hidden agenda... namely, the Wife has a great amount of difficulty these days 'hearing', especially when the 'dialog' is low, but the sound effects/music is loud, because when it gets too loud, she demands I turn down the volume... but then when the dialog is at some 'normal' or 'below' for some effect, she can't hear it.

I've noticed this problem on more recent BD movies we've been watching...

An alternative would be to send the audio through a device that I have easier access to the volume control and manually change things a bit more conveniently than getting up and turning a nob...
Another alternative may be to send the audio to a Bluetooth transmiter, and have her wear BT headphones...

As a note, note that... one of the disks I recently played had a 'late night' audio option... would that suggest a smaller dynamic range? So one could have the overall volume low, but not miss low dialog???
 
Yes you can (on every system I have seen) "calibrate" the speakers. So you could raise the center channel higher than would be normal. We don't have a room in the house where you could set up a surround system properly (turn of the century homes for some reason weren't set up to "watch" anything but out the window... But I have thought of getting a "surround" system and just using the LCR because the dialog will be clearer (even if you don't raise the level) if it doesn't have to share bandwidth with the explosions etc.
 
If the source material is digits, then you can raise the masters to give the right levels. If you are using analogue then raising the master, with the channels low will bring up the noise floor. I'm guessing the mix is within the editor, so should be fine.
 
So I'm playing back the edit on PPro CS6 and watching the levels which are averaging between -10 and -12db, but it does have peaks throughout close to 0db. Raising the master will bring those peaks over 0db, which I'm trying to avoid. This will mean having to re-adjust my levels to account for it and thus I end up with a different mix for web versus say projection in a real theater. But if that is what I have to do then that is fine, but wondering what is the best method for doing that? For web can I just take the master, bring up the levels and just compress it? That would be faster than adjusting every single individual track, no?

Cheers.
 
A peak-limiter is usually good to control occasional peaks. This process will alter the 'mix' some too, especially when used to excess.
 
mrl72 said:
So I'm playing back the edit on PPro CS6 and watching the levels which are averaging between -10 and -12db, but it does have peaks throughout close to 0db. Raising the master will bring those peaks over 0db, which I'm trying to avoid. This will mean having to re-adjust my levels to account for it and thus I end up with a different mix for web versus say projection in a real theater. But if that is what I have to do then that is fine, but wondering what is the best method for doing that? For web can I just take the master, bring up the levels and just compress it? That would be faster than adjusting every single individual track, no?

Cheers.
Click to expand...

Waves L2 or Massey L2007 will do the job just fine.
 
You must log in or register to reply here.
Back
Top