Does 1-pass cbr suffice at higher bitrates?

[Alvy]

Hello,
sorry this might have been discussed before but I did not find anything.
I do a lot of encodes from (analog) tv. because of the long encoding time I use 1-pass cbr with 1500 Kbps.
Is it highly recommented to use 2-pass even at this bitrates to achieve very good (not maximum) quality?

Another thing...
Something changed there with 1-pass cbr after 22-03-03. Now the encoder takes a lot of more bits into low complex scenes than before. now it achieves the desired bitrate (before absolutly not) ... but I preferred it the way it was before.
I tried get a simular result with changing cbr-options ... but without access. How can I make this mode "more vbr" ?

Thanks

Alvy (... another newbie)

[wuzel]

alvy,

i´m not sure whether i really do understand your problem, but i´ll try some kind of answer anyway.

you probably don´t intend to do 2-pass real-time encodings since this would be absolute nonsense as no-one can turn back time. that ability would be needed to do so (afaik).

if i get your point correctly you want to do a real-time encode at 1500kbs and then improve the material by doing another 2-pass encdoing just again at the very same bitrate.
this is not a good thing to do.

2-pass encoding is rather useful if you have a high-bitrate source that you want to shrink to a much smaller filesize preserving as much visual quality as possible.
(the typical situalition for DVD mpeg2 streams (approx 6 CDS for video only) you want to shrink to 2 or even only 1 CD.)

when the available bandwith gets reduced so greatly, a perfect distribution of the remaining smaller bandwith is critical for achieving maximum quality similar to the original movie.
(this is still true even though mpeg4 delivers better quality than mpeg2 at the same bitrate)

the improvement a two-pass encode can provide here is based just on the smarter share of the available bandwith after the source material has been analysed during the first pass.
this way bandwith is reduced at low-complex scenes that can be used in high-complex parts of the movie instead when it really isneeded.

of course the overall complexity of a movie can only be judged correctly after the whole movie has beed analyzed (with the fist encoding-pass).

if your original material is of rather low quality already (as it is the case with your real-time captured source) and you even want to keep the original´s filesize too, two-pass encoding doesn´t improve anything at all:

bandwith in low-complex scenes could of course still be reduced so it should be available for the higher-complex ones (theoretically). but since the original source is already limited to a maximum bandwith of 1500 kbs, the "surplus" bandwith you recieve in low-complex scenes can´t be added anywhere.
those excess details you could see with the additional bandwith you (theoretically) gained have already been lost when the source material was encoded at 1500 kbs.

secondly analogue tv-capture is not of the same quality as dvd-mpeg2 streams (well, some cards are better some are worse (8-bit/9-bit encoding etc). besides the analogue signal is distributed via cable or antenna so you are likely to have some enterferences or noise as well.

as a result i´d say: the quality you get with real-time 1-pass cbr encoding is not ideal, but good enough for analogue capture. do 1 cd or 2 cd or 1/4 dvd encodings, just as you need.

of course it is still possible to do 2-pass encodings with the material you captured.
but this is only sensible if you want to reduce your final filesize greatly.
so capture at extremely high bitrates - but you´ll probably run into codec saturation then or use losless compression for your original material (will need a very large harddrive)

i personally don´t believe the effort (mainly time) needed to do so will be worth for the results you´ll get compared with real-time 1-pass cbr, but this could be just a matter of taste either.

concerning your second question:

sorry, but i don´t believe that anybody will be able to help you out here.
you didn´t tell about the settings you used or which options you altered.

by changing quantizers for example you can easily mess up any capture.
if you did so - load default settings again. and save your default capture settings afterwards if you use vdub which i assume here.

afaik hitting desired bitrates in 1-pass mode has never been a critical problem (as long as packed bitsream mode is turned off)
i personally think that the quality problem you describe is gone with recent builds even though i have to admit i didn´t notice such a problem ever existed.

[Asmodian]

I am not sure what you are actually doing (as wuzel) though I will also try to answer but using a different assumption. I am assuming that you are encoding a 'captured from TV' video but are not actually capturing the video now.

I would strongly suggest that you do a two pass encode, at least as a test, and compare that to the same video done with 1-pass cbr. This is if you don't try any of the rest of this, at least see what you lose by only doing one pass. One option to see if you would gain much switching to 2pass from a 1500Kbps cbr encode is to play the encode with ffdshow and turn on the OSD displaying the quantizer. If it is over 4 with any regularity then you really might be losing a lot by not doing a 2pass encode.

Are you doing any filtering? What are you encoding in? Which version of Xvid are you using? The new versions of Xvid have some options that really help quality but at the expense of speed (and probably have better quality with those options off too). If you want better quality you could also try some denoising filters in avisynth (ie PixieDust or PeachSmoother). If this is captured from TV it is interlaced right? If yes, you would gain quality by doing a deinterlace such as smoothdeinterlace or fielddeinterlace (also in avisynth).
Another good option for quality (but with no size control) is the constant quantizer mode set to 2 or 3 (or 2.5 :) ).

[Zhnujm]

i have done a lot of encodes in cbr mode with rates from 1500-2500 (2cd encodes), but from digital captures. the quality was alwas good enough for me, but you can get surely better results with a 2 pass encode.
i would suggest you to just try it with one movie and compare the results (especially in fast motion scenes).

[avih]

i also used 1-pass CBR quite a lot. if you aim at high bitrates, try to enlarge the values in "CBR Options" if you're using xvid.

[Alvy]

Hello,
Thanks for yor answers. Sorry for my mistakable description.
Well I capture lossless with huffyuv (PAL 768x576). then I use vdub with temporal cleaner (up to 70) and 2d cleaner (8-12) optimized (and delogo 1.3) (ofter very poor signal :( ). Sometimes the source is interlaced, I use dynamic field order correction for that.

Xvid 14-05-03:
1 pass cbr
1500 kbps
h.263 or mpeg (depends on source quality)
bframes (3,130,100) threshold 80
chroma motion and vhq 1


times before:
Xvid 22-03-03:
1 pass cbr
3000 kbps (endsize was simular to new setting!)
h.263 or mpeg (depends on source quality)
bframes (3,130,100) threshold 80
chroma motion and vhq 1


Hmm, I checked some of my last encodes and there was the quant over 4 with regularity.
I already enlarged the cbr setting right now I do an encode with standard-settings (all of them) x10 (!)

The old version did not take much space on low motion scenes to keep source noise with full bandwith. But now it does - still with enlarged cbr - settings.