version 10 MPEG2 VBR file sizes

Chienworks wrote on 12/7/2010, 6:16 PM
Anybody else noticed sorta random file sizes when using VBR MPEG2 encoding?

I've been happily using the 600/minutes formula for DVD renders for years, always worked out fine. Now with Vegas 10 the file size seems to come out anywhere from 15% low to 25% too big. In fact, several times choosing a lower average bitrate has resulted in a large file than a higher bitrate.

Ugh. Video track copied & pasted into Vegas 9 for the render.

Comments

Chienworks wrote on 12/8/2010, 6:02 AM
VBR render complete in version 9, resulting file size was within 0.2% of expected.
Steve Mann wrote on 12/8/2010, 6:24 AM
Random as in the same project encodes different file sizes depending on the phase of the moon, or random that your 600/min formula is no longer predictive? (And where did you find that?)

If the former, then I have no clue. If the latter, then maybe V10 is using more efficient codecs. You can't reliably predict file size when encoding a VBR file. The V in VBR is "Variable", after all.
Chienworks wrote on 12/8/2010, 6:36 AM
That would be the first one, though that also causes the second to be true.

Do a forum search for "600" and you'll find hundreds of discussions about estimating output size to fit on a DVD. The general rule has been to divide 600 by the length of the program in minutes to get a Mbps value to use for the average bitrate. And, yes, V does stand for "variable", that doesn't mean it's unpredictable. Most MPEG encoders will create a file very close to what would be expected if the average rate had been used as a constant rate. As i indicated, Vegas 9.0e produced a file within 0.2% of the expected size, as it always has, and as have previous versions.

If Vegas 10 is using more efficient codecs then i would expect better quality in the same file size, but specifying an average bitrate should still define the output size to a pretty fair degree accuracy. If Vegas 10 were to pop up a window with a dancing, talking paperclip saying "Hey, i see you want to use an average bitrate of 7,000,000. Did you know that i can now produce the same quality you were expecting to see using only 5,783,191? Would you like to try that instead?", well, first i'd shoot the window and kill the paperclip, i suppose. But the point is that regardless of how well the encode is done, i want to fill the disc as close to full as possible to get the maximum quality that will fit on the disc. When i say 7Mbps then the file should contain 7Mbps and Vegas 10 isn't doing that. It's mostly making files larger by some random value, though sometimes they're a bit smaller. Having a better file doesn't help if it won't fit on the disc!
johnmeyer wrote on 12/8/2010, 7:46 AM
Kelly,

Something I stumbled into a few years ago may help. I was doing bitrate tests and found that the single-pass VBR in Vegas does NOT produce predictable results: you end up with file sizes that are much larger or smaller than what the average bitrate would predict. By contrast, the CBR and the 2-pass VBR produced predictable results.

So the question is: were you using single-pass VBR when your tests failed?


Chienworks wrote on 12/8/2010, 8:09 AM
Good thought, but no. Definitely double pass. I can see Vegas go through the video twice in the preview window, and the output file doesn't start growing until the second time through.
johnmeyer wrote on 12/8/2010, 8:28 AM
OK, that isn't it. Too bad, because I know for certain that single-pass has a problem.

Here's another idea, although unlike my last post, this one is speculation, and not based on my own experience.

I noted that the maximum bitrate default is now something like 9,800,000. This is much larger than the default max btirate we used to see which was something like 8,200,000.

I never thought this higher bitrate was legal, given the maximums specified in the DVD spec, so I always reduce it. Perhaps you could try again, but this time with the maximum set to 8,200,000. This is a long shot, and perhaps you have already reduced the max.