B-frames: about multiplicator,,,

[Gannjunior]

Hi,
I've tried to modify B-frames' mult from o.s. registry. I've tested in the 2pass : I've tried mult=1.00, 1.25 & mult=2.75, 3.00 and obviously the default mult=2.00 .The result is 5 identical avi.My friend made me note that,even if there's no differences at the end,the 2pass1st is affected by mult's changements: you can see it by differents values in the complexity.However in the 2pass2nd, the codec ignores changements in divx.log and does what it wants.
So I'd like to know if it's possible really changes the multiplicator.Besides, if I do a test with 2pass_10.000_q=Q=2+ Be,and I looked inside analyse.log, why all quantizers are riported=2?Since there are B-frames,because of mult,the avg quant. could not be 2...:confused:

thanks

ciao

[Acaila]

The best way to see the effect of the B-frame multiplier is by using 1-pass 100% quality. If you use 2-pass the codec will compensate for any differences and you will end up with equal sized files.

B-frames are determined at the 2nd pass, but the logfile won't show whether a frame becomes a P and a B frame. The info for a frame will be written to the log before the decision is made whether or not that frame becomes a B-frame (and hence will get its quantizer doubled).
I think anyway.

[Gannjunior]

My first attempts were own with 1passQ100 but differents multiplicators didn't affect the size. Since 1pass doesn't generate any log I decided to do the same tests in 2pass,hoping to learn something more froma analyse.log, but without any results...

If I want to count the avg quant.,B-frames on,it's enough I count all B-frames, in the considered film, and multiply them by 2 and after I do the average? B-mult = 2, really means X*2 ? Or means B-mult <=2 where the codec,according to the complexity about the single B-frames, decides the best "mult" to assigne to the considered B-frames...?
thanks

ciao

[Acaila]

B-frame multiplier = 2 means each frame that gets turned into a B-frame will get twice the quantizer than if it would have stayed a P-frame (in your words that is X*2). It should be able to change this value. But I don't know of anyone that has tried to change this (anyone out there reading this? :)).

Why you don't see an effect of changing the multiplier puzzles me. I know from XviD that changing the multiplier doesn't have an effect UNLESS you use 100%, 150%, 200% etc. Nothing in between those values alters anything (125% equals 100%, 175% equals 150%, etc.). Maybe it works the same way for DivX5?

[Gannjunior]

In Xvid,the change of "B-frame quantizer ratio", produces changements: in fact I obtained 3 differents files following the logic of multiplier: 300 < 200 < 100. The avi with multiplier=100 was the same to xvid with no B-frames.
As you can see from my previous post,for divx, I use also extreme values for multiplier (1.00 (-> 4096) & 3.00 (-> 11264) ) but with any changement..:confused:

ciao

[Acaila]

in fact I obtained 3 differents files following the logic of multiplier: 300 < 200 < 100.That's what I said, 100 works, 150 works, 200 works, but 125 does not.

As for why it doesn't work with DivX5 I have no idea, sorry.

[Gannjunior]

Thanks 4 your answers and if you find how to change the divx multiplier with concrete result post me.

ciao ! :D

[yingx2]

I tried to tweak bidir quant multiplier in registry, but failed, too.

Changing the value to 4096 (quant=1) or 16384 (quant=4) doesn't make any difference to my 100% quality-based encodes--file size didn't get any bigger or smaller.

Actually, i've got an idea about making bframes even worse than they are now, because i believe that it's useful for low bitrate encoding.

DivX 5 has the ability to "mask" those ugly bframes and emphazise only beautiful pframes during normal playback, am I right?

They really should make bframe quant switchable in future releases.

[Gannjunior]

Yes, you're right: B-frames use an intelligent way to lose informations: but in normal playback you're not able to realize the little loss of quality of the frames.That's true for quantizer's values enough low. In fact B-frames uses I&P frames but they doubles their quantizers; so,with the growing of quantizers of I&P,because of multiplicator,the "negative effect" of B-frame on quality is always more probable to see in normal playback.

ciao