Undersized with Bframes

[bienemaja]

I have a problem with undersized 1cd rips, while using bframes gives me always about 100megs undersized clips, leaving them away solves the problem. But i want to use Bframes cause they really help the quality of the movie clip.

Here are my settings

[.avs file]

I dont remember all filenames and i am not on my home computer so i cut it a little bit down.

Loadplugin="Mpeg2dec3.dll" (nics version 1.10)
Mpeg2Source="rd.d2v"
Crop(thecropsizeshere:)
LanczosResize(640,352) (i forgot if its 640x272 or 352)

[Settings]

Profile = AS @ L5
QT = MPEG (also tried H264)
Adaptive Q = on (already tried off here too)
Interlacing = off
Quarterpel = on (already tried off here too)
GMC = off
BVOPS = 2 / 1.50 / 1.00 (i already tried 0 / 1.50 / 1.00 or 0.75 here which helps)

Encoding Type = 2Pass (i left the advanced settings here by default)

MSP= 6 Ultra High
VHQ Mode = 4 (also tried 1)
FDR= 0
Maximum Ifame interval = 250 (for 25fps source)
Cartoon mode = off

Quantisizers = defaults
Trellis = on (already tried off here too)
debug tab = defaults

The target size is something with 603100 (i have about 90megs audio)

I also tried the Chroma Optimizer, but no changes then.

Oh before i forget it, the movie is Reservoir Dogs, but i had the same problems with Serendipity.

I already read the forum topics before about how to prevent undersized videos, but the suggestions there didnt help, except for leaving the Bframes completly away which i dont want to do.

So do you have any further suggestions for me ?

[buzzqw]

just a clue

and, ever tryed a "Compression Test" ?
Try with autogk or the big brother GordianKnot

BHH

[Imperial Llama]

Using b-frames correctly will nearly always lower the maximum file size. I'd strongly suggest disabling them (or reducing the amount used) if your rip is undersized because this will usually improve the quality as well as increasing the file size. B-frames help a lot for hard to compress sources, but when you source is so compressible that you end up with undersized files, they usually end up lowering the quality instead.

If you still want to use b-frames try increasing the resolution to 704*xxx or alternatively you could use the extra space to increase the audio quality.

[bienemaja]

I am pretty new to XviD, or better i never experienced this problem before with the alphas. How can i find out how many bframes i need ?

[bienemaja]

sorry for the reply on my own post...but does anyone have an answer on that ?

[mf]

Originally posted by Imperial Llama
Using b-frames correctly will nearly always lower the maximum file size. I'd strongly suggest disabling them (or reducing the amount used) if your rip is undersized because this will usually improve the quality as well as increasing the file size. B-frames help a lot for hard to compress sources, but when you source is so compressible that you end up with undersized files, they usually end up lowering the quality instead.
For the millionth time, disabling B-frames is BS. Just set the max count to 2, and ratio to 100. IS IT THAT $%^&*@ HARD?! If you don't know what to recommend, don't recommend at all!

[Teegedeck]

Except for the "BS [...]$%^&*@" I strongly second that.

And except for the 'don't recommend if you don't know [as much as I do...]'.

And except for the overall tone.

mf, come on...

[Didée]

Originally posted by Teegedeck
And except for the overall tone.
mf, come on... Perhaps the presents weren't like expected ?

:D

[mf]

Ok, sorry. I kind of flipped when I read another reply suggesting the wrong thing :rolleyes:. Maybe it should be put in a sticky somewhere. My apologies to Imperial Llama.

[Teegedeck]

@Imperial Llama:
buzzqw gave you hints, you seem to have ignored them. Do a compressiblity test. And search for 'saturated' on this forum and read the according threads; that should help you a lot.

[Imperial Llama]

@all

I have no bad feeling about what has been said so far. If I am giving out bad advice I want to be corrected (although perhaps a little less forcefully then mf did :D) and I would accept that the wording of my original post made it look like absolute fact, rather then just an opinion. However I would like to provide the reasoning behind my suggestion so it may be corrected if it is indeed wrong:

I thought that using b-frames always resulted in a slight loss of quality but that in a normal 2-pass scenario this is more than compensated for by overall reduction in size. When I have compared two clips encoded at Q2, one with b-frames and one with out, the b-frame clip is often slightly lower in quality. Of course the b-frame clip is also quite a bit smaller in size too.

Considering that bienemaja is already using no filters, lanczos resize and MPEG quant; he is already doing most of things that are usually suggested to help with undersized files. If I am correct about how b-frames help improve the quality then is this not the kind circumstances where the use of b-frames could be harming the quality?

Also I'm not sure how a compressibility test would help here. Obviously it will save him from encoding the entire movie to see if the changes he makes will help increase the size, but I don't see how it will tell him what changes to make in the first place. For example if he reduced the maximum key frame interval to 1 second it'd probably increase the size greatly but I don't think may people here would recommend this. (In fact no one here agrees with me on b-frames even though this would defiantly increase the size ;))

For the record, I always use b-frames for my rips because in my tests they have always improved the quality. If it's undersized I would increase the resolution, try different quant, etc before I'd consider disabling the b-frames, but if I was in bienemaja's situation I would be running out of ideas of what to do.

I be greatful if someone could comment on what I've said about b-frames. If they never harm the quality (or atleast not to an amount that could be detected by the human eye) then I'd like to know. I can provide samples of where I feel b-frames have harmed quality if this would be helpful to anyone.

[gldblade]

This is a complete guess, so please don't yell at me :D

I thought that using b-frames always resulted in a slight loss of quality This slight loss in quality comes from the fact that you are using higher quantizers for b-frames. However, using mf's suggestion of a ratio of 1.00, b-frames would be using the same quantizers as p-frames. As such, b-frames should have the same or better quality as p-frames. This potentially better quality comes from the fact that b-frames reference both the frames before it and after it.

Thus by leaving b-frames on with a ratio of 1.00, you may not only be decreasing filesize (although very slightly), you may also be increasing the quality of the video.

[mf]

Originally posted by gldblade
This slight loss in quality comes from the fact that you are using higher quantizers for b-frames. However, using mf's suggestion of a ratio of 1.00, b-frames would be using the same quantizers as p-frames. As such, b-frames should have the same or better quality as p-frames. I mention potentially better quality since b-frames reference both the frames before it and after it.

Thus by leaving b-frames on with a ratio of 1.00, you may not only be decreasing filesize (although very slightly), you may also be increasing the quality of the video.
Exactly.

[Imperial Llama]

That would make sense. I knew that b-frames were normally encoded at a higher quant but I didn't realise that encoding them at the same quant as p frames was preferable to disabling them all together. Thanks for explaining.

[Tommy Carrot]

I don't know, maybe i'm just unlucky, but giving the same quantizers for b-frames just increased the file-size in my experiments, and it wasn't any better quality-wise. Perhaps it's working for animes though, i tested only on natural contents.