bits/(pixel*frame)

[c0p0n]

I searched the forum, but I wasn't able to find a topic related with it (probablye because a bad search), so excuse me if this has been discussed here before.
A friend of mine, ReVolt, and me were talking about this matter from a couple of days.
I can remember that the recommendation of a b/pf between 0.2 and 0.25 was stablished for DivX 3.x. Since it is supposed that XviD is more efficient than DivX 3.x, would be true the supposition that b/pf could be undered to (let´s say) 0.18?
The main conclussion of this could be of being able to use a little higher resolutions to our XviD (and DivX 5) encodes gaining on overall quality and crispness of the image on fullscreen.
What do U think? is this a bullshit?

[kvs]

They should be more efficient, yes. However,
although you may already know this in
which case ignore, b/pp is IMO pretty much
BS as the only thing that counts is the source
compressibility. I mean, my encodes usually
have that number between 0.3 and 0.6, and
every now and then some anime wants up to
like 0.8 - you figure out the math if
you will, but it doesn't take a genius
to see why b/pp would be very stupid unless
you really have no other choice, of course.

But yes, both xvid & divx5 should require fewer
bits per pixel to achieve the same quality
as divx3, in most cases anyway ;) and some
will always disagree of corse!

[grug2k]

I think in one of the guides its mentioned that with XviD and Divx5 its safe to go 10-20% under.

So say...0.16 would be the lowest. I probably wouldn't go below 0.17 though, but as always, depends on what you're encoding. :)

[c0p0n]

yes, it is obviously on the gknot guide. I want to make some tests on it, as soon as I get time to (because of the final season).
you can make a compressibility test using the stats file of xvid (or to analyze a complete stat file), but the point now is what numbers do we have to look then on the compressibility test. I´m sure that this has been discussed by someone first, so I´m going to do some searches on the forum.
thanks for your quick responses!

[kvs]

As my own experimentation is often more mathematical than practical, I am not the best person to suggest settings, but as you don't seem to be aiming for the highest quality, I think 50% in the compressibility check in gordian know should do (it's only a percentage of first pass quant 2 filesize). 60% with xvid, and even more so with divx 5 (although it's all a bit of a cheat actually), should start approaching quite good, while 70% would be great. 80% would start to become overkill, and over 100% and you get undersized files obviously.

Now those values are related to the filesize of 1st pass, so 50% could give you b/pp number of 0.15 as well as 0.5 (if you're encode anime anyway harr harr harr). So there is no connection there, but for most liveaction movies, I think a percentage of 50% will give you quite a low b/pp.

And remember, it's usually better to have ie. a fair 576x256 encode than a poor 640x272 one ;)

If nothing of what I say is making sense, you need to read the boards a little more :) And if I'm only saying things you already know, sorry. I may have misunderstood your post.

EDIT: btw, if you're planning on low bitrate encodes, it would be even wiser to use the alternative curve compression, as it can do wonders there if you know what you're doing. :)

[c0p0n]

@kvs
but as you don't seem to be aiming for the highest quality
oh, no, no, you misunderstood me, mainly because of my poor english. I always aim for the highest quality possible to me. I asked this question precisely because of this point: I found myself using always resolution ranges from 480x*** to 576x***. In fact I do not use higher resolutions since that dark but beautiful flask 0.593 times.
my intention is to try to get a little higher resolutions to get more detailed images without loss of quality. this, obviously, depends on every movie, you cannot get the same quality on, let´s say 1 CD rip, Matrix and Hannibal without lowering the resolution of the last very much (or better, point to 2 CD). Or, best said, you are going not to get the same results on 1 CD because of the compressibility of Matrix is extreme (and on hannibal, apart from being a very complex movie relating to its images, you would have to lower too much the res).
on my encodes with xvid I always try to aim at least 65% on compressibility and 0.20 b/pf, but I´m begginnig to try to make a little testing on 60% and 0.17 as soon as I had a little time to. I think that the result should be reasonably better, but I could not assure this without doing something before.

ps: although I do not post too much because I have not as much time as I would like, I´m a member of the forum since ezboard times ;) and read it often.