is first pass always necessary

[kheops]

hi,

i apologize if this topic has always been discussed before, but it's pretty hard to do a "search" on it

well
i want to make a divx, using the two pass mode
i use a log and a mv file
let's say that once it's over the avi is only 450mb
really too short even if i add a 100mb mp3

so

can i just increase the bitrate and perform the second pass again ?
or do i have to start from the first pass again ?

many thanks

[khp]

According to Gej's divx encoding guidelines (http://www.divx.com/support/guides/guide.php?gid=1), you should not change the bitrate by more than 30% between first and second pass.

[yingx2]

I found something interesting in Gej's DivX Encoding Guidelines:

"The only parameter you can safely change is RC averaging period......the idea is to have the RC averaging period long as the half of the length of the clip you encode...... will allow more temporal variation of the bitrate and thus have a better handling of the fast motion scene. "

I never knew that. So, If i have a 120-minute movie at 25fps, I shuold set RC averaging period to 120x60x25/2 = 90000, shouldn't I?

I always leave that value as default--2000. Does Gknot automatically change that period for us?

[dragoman]

Hi,

No, Gknot sets it to the default, which is 2000.

If you want Vdub to have the RC of 1/2 the movie, you have to set it manually and do the two passes in Vdub.

Unfortuneatly Gknot does not offer this level of customization for it...

dragoman

[Fox Mulder]

Fortunately VdubConf 0.2 offer that level of customization for Gordian Knot:
http://forum.doom9.org/showthread.php?s=&threadid=25708

[theReal]

Does changing the RC averaging period make a big difference? Or is it just some kind of "if I want to see that it's better, then it is"? ;) :D

[llemor]

I have tried it before in DivX 5.0. When you changed RC averaging period to higher value(>2000), encoded avi has better quality, especially with fast motion movie. One drawback, you will not get your predicted size.

I haven't tested it with DivX 5.02. Maybe some other time.:)

[theReal]

One drawback, you will not get your predicted sizeAh, OK, a very major drawback. I wont change it on my next encodes then.

[yingx2]

Hi, all
My situation is just the opposite.

I encoded a 24-minute clip of A.I. with that RC averaging period customization, and compared it with the one that has the defult value.(2000, both 2-pass encoding)

I got nearly the same size and the same "bitrate distribution"--I split the encoded video into several shorter clips--the swimming pool scene, virtually static talking scenes, dark scenes..., and saw if that customization really allowed more bitrate for fast motion scenes.

I right clicked those files and selected their properties to view their bitrates(the data rate in explorer file property *8*1.024 equals the divx5 bitrate), but I didn't see any big difference.

The one with default value even gives slightly more bitrate(259KBps vs. 258KBps)to the swimming pool scene, which i think is a very hard-to-compress part.

Is there anyone willing to do a formal comparison to confirm this?
:confused:

I guess that we all want the codec to be smarter--fast motion scenes get more bitrate--even though it's already quite smart.( In my experiment, the bitrate of the swimming pool scene was twice, even thrice as much as that of the other scenes.) :)

[llemor]

@ yingX2:

What version of DivX did you use? 5.0? 5.02?

As I said my test is on DivX 5.0. Perhaps, I'll test it on DivX 5.02
;)

Thanks for the info.

[yingx2]

Sorry that I didn't mention it.

I use 5.02, as it's an apparently improved version.:)

[flooo]

i always thought that RC averaging period is for 1pass (not quality based) only encodings...

bla
flooo

[llemor]

@ yingx2:

I did testing using 20,000 frames for RC averaging period in DivX5.02 config. Change in predicted file size is negligible. No effect in speed. Quality-wise is not noticeable. Maybe I have to try with bigger values and really high motion movie.