Xvid & Target Bitrate?

[MaTTeR]

Is it possible to have a selction for "Target Bitrate" in the Internal 2ps tuning tab? I suppose it could be located under the "Target Filesize" field. I find this feature in Gknot to be most useful for testing purposes, trailers, etc.

A big thumbs up to all of you for todays latest build. I ran a 2ps ext/gknot test of "The Cell" trailer at a low 800 target bitrate. My jaw dropped when I seen the quality and compared it to the original VOB file. Many thanks for all your efforts:D

EDIT- If the target bitrate can't be added...would it be possible tp predict a bitrate beside of the target filesize field?

[-h]

I suppose it'd be possible, but it seems most people who perform 2-pass encodes start out by saying "I want this to fit into x CDs" - if you know how big you want the file already, it seemed the desired file size field was an easy way to skip a calculation.

Also, setting an integer "bitrate" will not allow file sizes as precise as the desired file size field.

It may be neater, for the codec to output a "calculated average bitrate = blah kbps" line in dbgview at the start of the 2nd pass.

-h

[MaTTeR]

Originally posted by -h
It may be neater, for the codec to output a "calculated average bitrate = blah kbps" line in dbgview at the start of the 2nd pass.

-h
Thanks for the quick response -h.

Your suggestion above would be more than adequate:) I suspect this would be much easier to implement. I'd be more than happy to help out, just let me know what you need.

[lar1r]

File size is useless for those who make non dvd rips. I tape a lot of tv shows and small clips of all different lengths. I want a constant quality using 2 pass.

Right now my best option is DIVX4, two pass. First pass I constrain the quantizers at 2. I then use a divxlog program on the log file and set my quantizer ranges. It then gives the exact bitrate for the quantizers I have set. This for me is the better solution.

I agree that for DVD rips, filesize length is better.

[MaTTeR]

@lar1r

Why not use Gknot with Xvid to accomplish this? Or am I missing something obvious here? ;)

[-h]

File size is useless for those who make non dvd rips. I tape a lot of tv shows and small clips of all different lengths. I want a constant quality using 2 pass.

Well, being able to set a bitrate will *not* give you constant quality, as each tv show will not be identically complex. DBZ would require more kbps than say barney the dinosaur, to look as good. If you want each clip to be the same quality, you would have to:

- Complete a first pass for every file
- Calculate the respective bitrate for every first pass (i.e. 1 file's first pass might have been 1300 kbps, another file might have been 800 kbps, but both would achieve the same quality of quant=2)
- Encode each respective file with an individual bitrate or file size, to fit into the desired space (1 CD, or whatever)

But just giving every clip the same bitrate, means some files will end up looking better or worse than others.

-h

[lar1r]

no no no - not the same bitrate, it changes based on the complexity recorded by the first pass.

The divxlog method kind of describes this: max all settings (ie:bitrate 6000, quant min/max at 2). Then after the first pass you can set a min max quant at lets say 2 and 7. Now the program will calculate a bitrate varied by the complexity recorded in the log and my new min/max levels.

This changes my bitrate for EACH file, gives me an accurate bitrate to use, and keeps the quality of each tv show between a 2 and 7 compression. Ex: My 1st tv show had a bitrate of 400, the second was 204, the third was 498. But each had the same boundries as defined by the quant min/max.

I understand that there could be a million ways to make a codec produce good quality. I just wanted to say that for my tv show recording purposes, I like to set boundries for the min max quants, and use the exact bitrate to achieve these boundries (as calculated by the program). Final file size is unknown to me because each file is different.

Gnot is a bastard to use. I have used it, and I don't like it. For a DVD movie though, I might be more incline to use it.

To tell you the truth, I would love to define my quantizer boundries at the start and let the whole 2 pass process run automatically. I HATE using more than one program to compress a video.

[-h]

Hmm you might want to try the "1-pass - quantizer" or "1-pass - quality" modes in XviD, they might be more to your liking. Plus you'd only be doing 1 pass :)

-h

[lar1r]

But nothing compresses quite like 2 pass... Plus I don't know what bitrate I need until i run the first pass to analyze the complexity and motion in my video.

Again - i just wanted to push across another point of view that some 2 pass users would not know the file size ahead of time.

Especially for animators who compress their own work.

[Alien9]

I'm one of the 2 pass users would not know the file size ahead of time.;)
I've video caputure card and I make my own uncopressed avi...
Files have different lengths so I'm not able to set standard final file size... Usually I use DIX4 - 2-pass and I set bitrate at 1400; now I want use XviD but whiteout "Target Bitrate" selection I can't :(

Alien9

[Doom9]

the whole point of 2 pass is to know the size in advance. What is bitrate other than size? bitrate * length = size.. it's really that simple.. divide size by length and you have bitrate. xvid is very accurate with size, as is nandub.. so in both you can set a certain size and get it.. divx4/5 actually has a flawed rate control.. eventually it will work the same

[trbarry]

My problem is that the length of my source varies more than the complexity or quality of my sources.

And I've been trained by Divx to have a feel for what bit rate gives me what results. Recently with Xvid I've been adjusting that down a bit (this is good!) but I still find myself checking the length of the film and multiplying it by the bit/byte rate to tell Xvid what to do. I'm not trying to fit them onto CD's but I still like the fact that 2-pass seems to allocate bits better to the places that most need them.

To be honest, I'd probably rather be specifying 2 pass quality as bits/pixel.

- Tom

[tangent]

Then use Gordian Knot, it gives you all the information you need.
But of course, the most useful check to determine how good your final output quality will be is the "% of 1st pass"