XviD-26012003-1 (unstable) developer build [Archive]

terwin

27th January 2003, 10:30

Yes, good work.

I have just testet a short clip (one chapter of Star Wars Eps2) with b-frames 2/100/100, chroma motion and qpel and it works fine for me. No oversize problem.

But XviD doesn't pay attention to the maximum i-frame intervall. I have set it to 300 frames but in some cases there are more then 600 frames between two keyframes.

cu terwin

ookzDVD

27th January 2003, 10:34

@Koepi,

Oversize problem is fixed. :)
Thank you for your great effort.

The Link

27th January 2003, 11:31

just want to confirm again: No filesize problem any more! I tested several combinations on a short clip and XviD has always hit filesize on the point!! :) :) :)

Regards,

The Link

NuclearFusi0n

27th January 2003, 12:40

Originally posted by terwin
Yes, good work.

I have just testet a short clip (one chapter of Star Wars Eps2) with b-frames 2/100/100, chroma motion and qpel and it works fine for me. No oversize problem.

But XviD doesn't pay attention to the maximum i-frame intervall. I have set it to 300 frames but in some cases there are more then 600 frames between two keyframes.

cu terwin
read the forum - it's known.

The Edge

27th January 2003, 13:14

I see over on betanews (http://fileforum.betanews.com/detail.php3?fid=1043648426), Koepi's 26012003-1 XviD gets a mention.
Also gets 5/5 :)....well deserved.

Very nice codec. In my opinion it's better than the DivX 5.03. I can make my files better looking and smaller with XviD. Keep up the exelent work XviD developers and fans :-)

Again, thanks Koepi.;)

Edge

Sigmatador

27th January 2003, 13:42

thx koepi your build rox again ^c^

what about the max iframe interval bug ? (can someone confirm terwin ?)

iago

27th January 2003, 13:46

@Sigmatador

Since it's a known problem, and since it's stated one more time even in this very thread, does it make any sense to repeat the same issue over and over?!

iago

Sigmatador

27th January 2003, 14:49

ok, forget what i wrote ;) (but it's the last bug to obtain a monster build :D )

iago

27th January 2003, 14:57

"...monster build..."

Hehe, I like it! :D

b1ahb1ah

27th January 2003, 18:30

thx Koepi!
You r0x and xvid too !:)
I just test with an extract from The Silence of the Lambs, and no oversized, very good job koepi ...

terwin

27th January 2003, 21:21

Since it's a known problem, and since it's stated one more time even in this very thread, does it make any sense to repeat the same issue over and over?! Ok, sorry. :o

But I have always read about a problem with the minimum i-frame intervall. Besides it wasn't mentioned in kilg0r3s bug list.

Foxer

28th January 2003, 01:05

Took a quick look at the sourcecode..

It turns out that the core is now setting something it was supposed to and the vfw had made reservations for it but for some reason, it was decided that the variable it was supposed to be stored in was to be used for B/SKIP/PAD/DELAY flags.

Since this change, rate control w/ bframes has been messed up.

Koepi, until the new frametype flag or hlength variable is decided upon, an easy fix to make in your build is to zero nns1.dd_v instead of setting it to stats->hlength in codec_2pass_update() in 2pass.c

NuclearFusi0n

28th January 2003, 01:22

OMFG the quality in this encode i just made - wow. wow!! XviD is progressing at such a fast rate, thank you Koepi! :)
now we just have to wait for the next version of ffdshow and we're all set :D

hayami

28th January 2003, 05:39

Just finished a 56min 2pass encode, filesize problem is gone now =)
ChromaME + LM + Bframe = rocks!

NuclearFusi0n

28th January 2003, 10:23

Originally posted by hayami
Just finished a 56min 2pass encode, filesize problem is gone now =)
ChromaME + LM + Bframe = rocks! Don't forget QPEL and GMC! :D

tiki4

28th January 2003, 14:31

Yess!

Encoded Wild Things on 1 CD yesterday: QPel + Chroma + GMC + B-frames looks better than ever (never used GMC before in XviD, but there seem to be some scenes where it is really useful).

:) :) :)

Keep up the great work, all of you!

tiki4

Boardlord

28th January 2003, 16:20

Maybe repetition, but just finished encoding Stripes with qpel, gmc, bframes, chromaME. AWESOME for me too!!!
If only we had an ffdshow build with the simple idct now...

kilg0r3

28th January 2003, 19:42

is there a way to manually change iframes to b/pframes in the statsfile?

NuclearFusi0n

28th January 2003, 22:36

http://iupload.net/012003/xviderror.png

Settings for XviD ()
Mode: 2 Pass - 2nd pass Int.
Dummy 2nd Pass: OFF
Desired Size: 1045477KB
I-frame Boost: 0%
Below I-frame Distance: 10%
I-frame Bitrate Reduction: 20%
Curve Compression: Payback with Bias, Bitrate Payback Delay: 250frames
High Bitrate Scenes: 0%, Low Bitrate Scenes: 0%
Hinted Motion Estimation: OFF
Alternative Curve System: OFF,
Motion Search Precision: 6 - Ultra
Quantization Type: MPEG
FourCC Used: XVID
Max I-frame Interval: 300, Min I-frame Interval: 1
Lumimasking: OFF
Interlacing: OFF
Greyscale: OFF
Quarterpel: ON
Global Motion Compensation: ON
Chroma Motion: ON
Max B-frames: 3
B-frames Quantizer Ratio: 150%
B-frames Quantizer Offset: 100
Packed Bitstream: OFF
DX50 B-VOP Compatibility: ON
Print debug info on each frame: OFF
Min I-frame Quantizer: 2, Max I-frame Quantizer: 31
Min P-frame Quantizer: 2, Max P-frame Quantizer: 31
Max Bitrate: 10000kbps, Max Overflow Improvement: %, Max Overflow Degradation: %
Start Credits: OFF, End Credits: OFF,

The first encode of the matrix went fine. I then made a new AVS file, tried encoding it, it always dies out on the 2nd pass. Here's the AVISynth script
LoadPlugin("C:\Program Files\AVISynth\mpeg2dec3.dll")
mpeg2source("F:\Matrix\Matrix.d2v")
crop(0,64,720,352)
Full anamorphic resolution, 720x352

MoonWalker

28th January 2003, 22:42

Hmm...I get this message when XviD fails to load the 1pass stats file..Are you sure the stats file is accesible for the 2ndpass??

MoonWalker

P.S Windowsblinds???

NuclearFusi0n

28th January 2003, 23:56

Originally posted by MoonWalker
Hmm...I get this message when XviD fails to load the 1pass stats file..Are you sure the stats file is accesible for the 2ndpass??

MoonWalker

P.S Windowsblinds???
well, it encoded all the way to the scene where they break morpheus out of the compound with the bigass helicopter, and it looked great all the way up to that point.

not windowblinds, XP themes ;)

MoonWalker

29th January 2003, 00:01

Hmm..Maybe your stats files got corrupted somehow?? Can you open it with GKnot or XviD Stats Viewer (http://homepages.pathfinder.gr/statsviewer) ??

MoonWalker

frodoontop

29th January 2003, 00:11

I had about the same problem with the second pass. There was a simple solution though. Just hit the load default button and restore your old settings(they didn't seem to matter) and things work again. At least for me. Hope this is helping...

NuclearFusi0n

29th January 2003, 00:22

Originally posted by frodoontop
I had about the same problem with the second pass. There was a simple solution though. Just hit the load default button and restore your old settings(they didn't seem to matter) and things work again. At least for me. Hope this is helping...
I load defaults every time i install a new build

what i find strange is that it worked FINE for encoding the matrix at 672 x 272, and when i use the EXACT SAME SETTING for a higher resolution encode it does this - makes no sense to me :( (i just changed the AVS file and ran the exact same job again in virtualdubmod's job control)

I'll try it again when my encode of Donnie Darko finishes up.
Originally posted by MoonWalker
Hmm..Maybe your stats files got corrupted somehow?? Can you open it with GKnot or XviD Stats Viewer (http://homepages.pathfinder.gr/statsviewer) ??

MoonWalker
I'm going to run through both passes again, it takes a decent 8 to 10 hours or so to encode on my Athlon XP 2000+. See ya tomorrow :D

SiXXGuNNZ

29th January 2003, 07:13

very nice, I encoded chapter 15 of the matrix with divx 3.11, divx 5.03 pro, koepi's 04102002 and koepi's 26012003

low motion on divx 3.11(it looked better then divx 5.03)

divx 5.03 I only used bi-directional

with 04102002 I used the settings from the guide on doom9

with 26012003 I also used the settings from doom9 but for b-frames I used 3/150/100 and chroma motion only

all 4 seemed to hit the filesizes pretty dead on.

I can barely tell any diff between the two xvid encodes, but the divx 5 was a 10 pass and looked like crap imo

the two xvid encodes were both 2pass

what's the eta on a stable binary with the b-frames enabled? I only ask because of the compatibility thing. personally I think the new build is rock solid, will try out qpel and gmc later.

NuclearFusi0n

29th January 2003, 08:55

OH MY MOTHER F-ING JEESUS CHRIST WOW!

I'm encoding movies on 2 cds, with FULL ANAMORPHIC RESOLUTION (minus the crop for 2.4:1 movies), and 5.1 448Kbps AC3 tracks and it's looks ABSOLUTELY FANTASTIC.

:eek: this is absolutely jaw dropping quality. from my viewing position, a few feet away from a 19" diamondtron monitor, it's VERY hard to tell the difference between the dvd and the encode. i'd have to be looking for artifacts. the average layperson (read: cute chicks that i wish would watch movies with me) couldn't tell the difference.

wow.

To every single XviD developer: THANK YOU!

Teegedeck

29th January 2003, 09:18

That about summs it up. Ugly two-pass hacks or not, in the not-too-distant future, I'll have to list the name of every single developer in a big :) thank you note here on the forum. Boy; B-frames, chromaME, GMC and Qpel! Every single one of them actually working unlike some well-known commercial codec! And instead of just relaxing the developers are thinking about how to tweak the MPEG4-standard to get even more quality-improvements out of it - boy, boy, boy; you guys are just great!:cool:

Garfield

29th January 2003, 11:58

I made a 2 pass encoding of the Big Lebowsky with B-Frames and Chroma Motion only, with linear curve by defauts and the agressive quantizers settings in iago's tutorial and i'm very impressed by the result.

Thanks to all developpers for this nice work

Sigmatador

29th January 2003, 11:59

[i]the developers are thinking about how to tweak the MPEG4-standard to get even more quality-improvements

I love it :D

The Edge

29th January 2003, 12:13

I totally aggree with everything said so far.....
For a though test for the new build I encoded LOTR:TTT dvd (yes, I'm a bold boy:eek: ) and what can I say.
It's a 1CD encode with an average bitrate just shy of 500kbs
XviD & mp3.....Quarterpel, GMC and CMotion all enabled.
Thanks XviD. I look forward to XViD 1.0

Edge

tiki4

29th January 2003, 12:33

@The Edge:

Maybe it's time now to think of how to get the bitrate of multichannel audio down to say ~64 kBit. If Monty at xiph just had time to implement some more things into Vorbis like full channel coupling...

What comes next? 3 hours of LOTR on 1 Floppy?

Cheers,

tiki4

The Edge

29th January 2003, 12:56

@tiki4
That would be tough. I remember reading that microsoft said it could be done with their super-duper windows media codec:)

Edge

tiki4

29th January 2003, 13:01

Back to topic:

I compared a first pass with the latest stable build of Koepi and with latest unstable with all features enabled (Chroma, QPel, B-frames 2, GMC). The size went down from something like 1100 to 700 MB. Maybe there will come a time we don't need a second pass anymore. Movie was still 'Wild Things'.

tiki4

iago

29th January 2003, 13:02

Hello everyone,

I also agree with all the praise that goes to this latest build by Koepi. Yeah, visual quality is absolutely striking, b-frames work great (only except Min-Max I-frame intervals in two-pass), file-size accuracy is re-gained, and even Qpel and GMC are behaving now! ;)

Thanks again to all XviD developers for this great achievement!

best regards,
iago

sam_b

29th January 2003, 14:58

This new build has got be be breaking the laws of thermodynamics or entropy or something :D Scary quality. Fastastic job all you devels.
But I like NuclearFusi0n's comment the best :)

Two questions though? How are you all viewing the output, ffdshow, xvid.ax, envivio, 3ivx.....? And, (sorry if this is old) is RefDivx's lumi code in CVS now? If not, what is the reason?

majerle

29th January 2003, 15:41

YAAB ! (Yet Another Awesome Build)
Truly fantastic, i encoded an old, noisy film "The Warriors".
As you can suppose 1 cd backup( Bf 3-150-200, GMC, CM, Qpel) ,ogg 96k, near full resolution (688*xxx) , and avisynth 2.5b(plus come c3d) : FANTASTIC!

Thanks guys ! Nothing more to ask.

Andres

EDIT : 690mb + zoomplayer3.5 + oggds0995+ ffdshow(latest) --> A near DVD

Nic

29th January 2003, 16:22

Just as something to do, I uploaded a new build with a new DShow filter.

The DShow may still have the same old problems, or it may be working ok. Its hard for me to say. Seems ok though :)

http://nic.dnsalias.com

-Nic

ps
Mine won't have Koepi's oversize fix.

bond

29th January 2003, 16:53

Originally posted by sam_b
...sorry if this is old, is RefDivx's lumi code in CVS now? If not, what is the reason? Originally posted by Nic
ps: Mine won't have Koepi's oversize fix. hm i hope that the future xvid development will not be divided too much :(

kayman

29th January 2003, 18:15

ive been encoding a few episodes of jeremiah from MPEG tranport streams and have had no problems till this build i do my first pass and have my 2nd pass set to "594121" to accumalte a 121 meg ac3 audio track but it only creates a 360 meg avi?

help!

:)

peaudepailles

29th January 2003, 18:36

Hi,

I had sometimes the same problem with older builds 24012003 and 03012003. (particulary with very dark movies) If you used bframes, try to disable it, thats help a lot (for me ...)

And you can also used MPEG quantization type if not already done,
and no resizing filter.

Hope this help.

Peaudepailles.

@Koepi : Thx in advance for the new build. Not tested yet but according to all, thats a great job !

Bravo ! (as we tell in france ;))

NuclearFusi0n

29th January 2003, 19:21

Originally posted by sam_b
This new build has got be be breaking the laws of thermodynamics or entropy or something :D Scary quality. Fastastic job all you devels.I honestly thought I oversized it, 2pass had a bug, or i had put in an extra zero in the size box. I couldn't believe my eyes. I had to double check the filesize, and make sure i was looking at the right file - i was amazed.
Originally posted by sam_b
How are you all viewing the output, ffdshow, xvid.ax, envivio, 3ivx.....?I'm using ffdshow in ZoomPlayer with "Use Xvid" checked ON (xvid.dll) because of qpel smearing, resize to my screen resolution, and aspect ratio correction for my Anamorphic encodes.

NuclearFusi0n

29th January 2003, 19:39

Originally posted by kayman
ive been encoding a few episodes of jeremiah from MPEG tranport streams and have had no problems till this build i do my first pass and have my 2nd pass set to "594121" to accumalte a 121 meg ac3 audio track but it only creates a 360 meg avi?

help!

:) that means it's "maxing out" the quality. try increasing the resolution (if you are decreasing it, but dont go higher than the original resolution) or try disabling bframes.
:)

Didée

29th January 2003, 23:14

Originally posted by NuclearFusi0n
[...] resize to my screen resolution, and aspect ratio correction for my Anamorphic encodes Seems so many people get fanatic about anamprphic encoding these days (see also several threads in the "New A/V Formats" forum ...).

LOL.

Anamorphic encoding is intended for 16:9 displays. Or, more precisely, for offering the best compromise when the target display AR is unknown.
See, we're usually compressing the video down to the bone, more or less. Where is the sense in wasting bits in additional resolution, introducing some additional artefacts, when you reduce the resolution again on playback anyway? Making the artefacts smaller, that you wouldn't have had without anamorphic?

Or are you making some kind of "backwards anamorphic" encoding (not raising Y, but reducing X) to lighten the overall resolution? If so, then please accept my apologies ;)

Didée

NuclearFusi0n

29th January 2003, 23:48

Originally posted by Didée
Seems so many people get fanatic about anamprphic encoding these days (see also several threads in the "New A/V Formats" forum ...).

LOL.

Anamorphic encoding is intended for 16:9 displays. Or, more precisely, for offering the best compromise when the target display AR is unknown.
See, we're usually compressing the video down to the bone, more or less. Where is the sense in wasting bits in additional resolution, introducing some additional artefacts, when you reduce the resolution again on playback anyway? Making the artefacts smaller, that you wouldn't have had without anamorphic?

Or are you making some kind of "backwards anamorphic" encoding (not raising Y, but reducing X) to lighten the overall resolution? If so, then please accept my apologies ;)

Didée
anamorphic encoding is intended to increase vertical resolution so that you dont waste resolution on "black bars."
"Or, more precisely, for offering the best compromise when the target display AR is unknown. " no, it's specifically designed to lower resolution "wastage" on widescreen material.

"Where is the sense in wasting bits in additional resolution, introducing some additional artefacts, when you reduce the resolution again on playback anyway?"
I'm not reducing resolution on playback.

think about this:
the actual video is 720 x 480
When resizing for proper aspect ratio, most users reduce vertical resolution (example, resize to 720 x 395)
what anamorphic encodes do, is stretch horizontally, to keep full vertical resolution (880 x 480 for example)
http://gregl.net/videophile/anamorphic.htm

example of the anamorpic difference: http://www.thedigitalbits.com/articles/anamorphic/anamorphic185demo.html

ookzDVD

30th January 2003, 05:06

Originally posted by Nic
Just as something to do, I uploaded a new build with a new DShow filter.

The DShow may still have the same old problems, or it may be working ok. Its hard for me to say. Seems ok though :)

http://nic.dnsalias.com

-Nic

ps
Mine won't have Koepi's oversize fix.

@Nic,

I can confirm that your DS Filter is still have problem while seeking.

Gazza

30th January 2003, 07:28

To all the developers and testers - the latest build is very very nice. I've always stuck with xvid and now there is absolutely no reason on earth to go elsewhere. Many thanks for a truly impressive effort.

Just a question - I'm interested in finding out if anyone still uses convolution3d, filters, what resize method, etc? I'm wondering if using these is actually going to remove quality from what xvid is doing now?

Teegedeck

30th January 2003, 09:31

Well... Every attempt at filtering will make the output look less similar to the original. And XviD compresses very well with little side-effects. So there are IMHO only two reasons that justify filtering:

1) Increase compressiblity: You're not able to put it onto two CDs otherwise or not willing to use more than 1 CD for that movie
2) The original source looks like shite (Tv-capture) so you don't have to do it just for compressibility but also to make it watchable

As -h said, noisefiltering generally helps MPEG4, so theoretically it's a good idea to use as much as is possible without a subjective reduction of quality. Problem is, if you're a cineast, you won't want to soften a superbly looking picture at all. So you have to compare what looks better to you; strongly quantized, unfiltered image; less quantized de-noised image; filtered or unfiltered anamorphous image; or smaller resolution with less quantization. For my personal taste, if I have to filter I prefer MPEG2DEC3's postprocessing with everything except for luminance deringing (too strong softening effect) to get the size down and perhaps fluxsmooth if it's still to large. Smaller res looks better than a picture that's too soft. Also I feel that clean sources may look better if you keep them anamorphous (resized if neccessary - I go for Lanczos) but compressibility has to be above 50%, so it's still a rare occasion... ;)

I guess I'm just repeating stuff that isn't new to anyone here, sorry. To put it short: XviD works pretty well on it's own; I wouldn't use filtering if I don't absolutely have to. Hope you find that yabber a useful in some way.