PDA

View Full Version : divx5(.1.1) first pass size calculation - how ?

len0x
3rd December 2003, 19:44
I'm trying to do a very simple thing - find out the first pass size of the multipass encode. Unfortunately DivX don't output AVI during first pass, so the only way to get some information is log file.

Right, we have total_bits field in there and I would have hoped that I can just sum it up, but it turned out that that field is not the same as quant 2 single pass encoding (as I was hoping). So in fact it not the maximum quality, but somewhat 30-40% less than that.

So the problem is - I have compressibility test value, which I cannot confirm after the first pass :( Any solution to this I wonder?
temporance
3rd December 2003, 20:49
Why don't you just run a Q=2 first pass?
len0x
3rd December 2003, 21:06
The same reason I'm running comp test - time consideration. It's enough to have comp test value in the beginning, but I just want to confirm how close it is to the actual first pass size percentage.
(not that I want to adjust anything, just to see the value for information purposes)
temporance
3rd December 2003, 21:37
Maybe you can trick the first pass to do a Q=2 encode by setting a ridiculously high bitrate? Check C:\FrameLevelControl.txt to see what quant distribution you get.
DigitAl56K
4th December 2003, 11:28
You can run a q=2 encode using 1-Pass, Quality Based mode. This mode actually outputs a log file that you can use for Multipass, Nth pass.

Also, turn off any PVE when doing the compresability test.
cordraconis
4th December 2003, 12:57
Originally posted by DigitAl56K
You can run a q=2 encode using 1-Pass, Quality Based mode. This mode actually outputs a log file that you can use for Multipass, Nth pass.

Also, turn off any PVE when doing the compresability test.
:eek:
So, since I do a full Q2-pass for my compressibility check, (And as an indication of filter/resizing behaviour), I could use the log file for my first n-th pass and skip the whole first pass?
This would save me at least half a day!

But how is this influenced by the slow or slowest modes for my nth-passes? or PVE?, or the bitrate?
:confused:

Could you please give some more info on the interchangeability of that Q2 logfile?

TIA