Codec War: DivX vs XviD

[SoundScape]

Hello,

I've been using DivX Codec for quite some time now and I can say I am happy with the results I get except for one thing: very smoothed video, especially at low bitrates. I have tried other MPEG-4 codecs like HDX4 and VP70 but these too performed worst or in some cases equal to DivX (especially VP70). Some of you might wonder if I have ever heard of XviD :) Of course I have! This is exactly the reason I am posting this as I need some advise on XviD. I've been a die hard fan of DivX and because of that it was hard for me to turn my attention to other codecs that might give better results. One day somebody told me that XviD is said to keep much more detail than DivX. I was sceptical about that and as I said before I didn't want to experiment with another codec, especially XviD for a lot of reasons. One of them was that XviD was constantly under development (beta, RC releases) and I didn't like the idea of doing my encodings with so many different versions of the codec. Stupid thinking? Maybe...
Of course I was closely observing the developement of XviD and after a lot of reading on Doom9's forums I decided to give it a try. Below there is a link with 3 images taken from 3 different video sources. The original DVD video, DivX and XviD. DivX codec version used is 5.2.1 Pro while XviD version is Koepi's build 1.1.0-Beta1-16012005. Oops, did I just used a beta release of XviD despite of everything I told you about not liking beta releases? I know, it doesn't make much sense but the reason I used it is because I wanted to test the latest release of the codec. Anyway, in both encodings a target birate of 800kbps was used so as to test the performance of the 2 codecs in low bitrates. DivX was configured as follows: 1st pass - Encoding Performance "Standard", Psychovisual Enhancements "Fast" and 2nd pass - Performance "Slow", Psychovisual Enhancements "Fast". XviD on the other hand was left at its default settings (don't want to mess with so many settings, right?).
You can view the results here (http://www.picjar.com/pub/SoundScape/DivX_vs_XviD/). I believe that BOTH codecs produced a very smoothed video (facial details are wiped out) and that what I've been told about Xvid's capability to retain more detail than DivX, does not seem to apply here. Maybe I did something wrong or XviD needs to be configured in a specific way. To be blunt, I didn't see any major differences between the 2 codecs and I still don't see any good reason to stop using DivX and switch to XviD. Of course no one forces me to do that! I am just sharing my thoughts with you, that's all. I believe that XviD is a highly configurable codec and if one knows what he is doing, great results can be achieved with it. Even better than DivX. I am not one of them. yet.
Thank you very much for your attention and please do feel free to post your own thoughts on the above. Your opinion is of great value to me.

Thank you!

[jon.schaffer]

Frames with exact same cropping would have been perfect ;)
You did not say the achieved bitrate...

Look closely at the hair of the guy: they are sharper with XviD...

EDIT: should we talk about a... a war?

[dragongodz]

SoundScape - to compare the pictures load them in a viewer where you can flip back and forth. or even just load the pics in that link on different pages and then flip back and forth.
look at the hair, look at the handle and lock above his head, look at the mark on the letter on the left, look at the hair(as jon.schaffer said) etc etc etc. xvid doesnt look a bit sharper to you ?

[*.mp4 guy]

Turn on Vhq for bframes too and set vhq to 4 for better quality (you used divx on slow so speed is unimportant right?) then read and Xvid guide and choose the settings you like best. Fyi more vhq is always good, but slows down encoding a bit.

[Andrey]

Bright example of DiVX smoothing a lot ;)

[Sagittaire]

1) If you want really compare codec frame comparison is impossible and doesn't prove anything: DivX use perhabs bframe and XviD not. Conclusion for N and N+1 Frame are perhabs different.

2) DivX 5.2.1 is not the last and best DivX codec: Use DivX fusion with Unconstrained, Insame or Extreme quality, Multibframe, H263 optimized, and without psy, GMC or Qpel
http://labs.divx.com/archives/TESTFusion.pdf

Here (http://forum.doom9.org/showthread.php?s=&threadid=90784) my codec test with metric and trailers for visual comparison ...

[Andrey]

>>and without psy, GMC or Qpel
Why now Qpel ? Any problems ?

[Sagittaire]

Qpel activation is impossible with RD optimisation (best, extreme and insame) for DivX Fusion.
GMC is not optimized (better metric without GMC)
Psy is not optimized (very better metric without Psy, visually smooth with Psy)

[IgorC]

There is no any war.
Sagittarie, you don´t use latest version of Xvid beta1 which is better.
Also you don´t test rev178 x264 wich is better than VP7 for SSIM test . Maybe you were comparing VP7 and x264 at the moment when VP7 official beta was realesing (at this point x264 had rev157). Howewer there is no point to compare like this. And you didn´t answer why you used not best settings for x264. From this point of view I don´t see the reason to believe your ´test´.

[Andrey]

Oh :)
Thanks !
Their enchantments always did a picture smoother before, so I tried to avoid them (RD too).
Need to refresh DiVX knowledge ASAP :cool:
EDIT: a bit.

[Andrey]

Sagittarie, you don´t use latest version of Xvid beta1 which is better.
XViD 1.1 is beta and still have problems. It will be compared after release, I hope.
From this point of view I don´t see the reason to believe your ´test´.
Then make your own. Point to someone errors and bad attitude are different things, IMHO.

[IgorC]

Andrey. There is test between betas (Divx Fusiob beta2 , Xvid 1.1 beta 1 march 14). Please read the test. At least Xvid haven´t problem with PSY and GMC.
I´m agree with result of OPSNR test but visually Divx´s video is so smooth (without PSY too). How can you talk about Divx as best ASP when it´s not realsed yet? Ok I can also say that XVID 2.0 will be better than Divx6. But Xvid 2.0 isn´t realsed yet. don´t you see?

[IgorC]

Originally posted by Andrey
Point to someone errors and bad attitude are different things, IMHO.

I and other people pointed errors (like custom matrix and other settings ) but there wasn´t answer for it. ignorance?

[Andrey]

>>There is test between betas
DiVX betas looks much as a release, with some features cut. But I get your point. Agree.

>>At least Xvid haven t problem with PSY and GMC.
GMC and DiVX are incompliant things :D

>>I m agree with result of OPSNR test but visually Divx s video is so
>>smooth (without PSY too).
Yeah. That is. And always was. I know :p

>>I and other people pointed errors (like custom matrix and other
>>settings ) but there wasn t answer for it. ignorance?
Could be :) But this is not the reason to strike back ;)

[Andrey]

Oh, I forgot about original poster.
SoundScape:
and that what I've been told about Xvid's capability to retain more detail than DivX,
IMHO it is applied not to any case, but for rather high ~1 - 1.5Mbit bitrates. It is becasue in XViD you can select MPEG matrix (the same as for MPEG2 I think) which does not smooth details as h263 scheme do.
h263 scheme (matrix with all coeficients set to 2) was intended to be used for small size videos on low bitrates, when block size became comparable to the picture size. So it's results can be disaponting on D2 resolutions...
EDIT:
XViD with h263 has often more details than DiVX too - because of advanced features it uses (like Qpel, psy enchancements and so on)

[IgorC]

You can also select MPEG matrix for Divx but H263+ is best matrix for Divx at low and middle bitrates (1CD and 2 CD rip).

[Sagittaire]

Originally posted by IgorC
There is no any war.
Sagittarie, you don´t use latest version of Xvid beta1 which is better.
Also you don´t test rev178 x264 wich is better than VP7 for SSIM test . Maybe you were comparing VP7 and x264 at the moment when VP7 official beta was realesing (at this point x264 had rev157). Howewer there is no point to compare like this. And you didn´t answer why you used not best settings for x264. From this point of view I don´t see the reason to believe your ´test´.

Between rev 158 and rev 178 there are only cosmetic update ... 178 isn't very better than 158. Between Xvid 1.1 beta 1 march 14 and XviD koepi 1.1 beta 1 too ...

But send your encoding if you can make better encoding ... :D

[IgorC]

I delete file of encoded video a few days ago. But try these settings
rev157 -> rev178 Cosmetic? I would say cosmic difference. http://mirror01.x264.nl/x264/changelog.txt

x264(your trailer) rev157 SSIM=68.24 , OPSNR = 40.4423
VP7 (your trailer) SSIM=68.57 , OPSNR = 40.6802
x264 rev174 SSIM=68.77 , OPSNR = 40.4737

settings for x264 rev174 - Cabac, Deblock -1, Bframes 2 (Temp), Weight, ref15, keyboost 50% , Breduc 40%, VBR 75%, 5 max Quality , macroblock partitions ALL on. I don´t´know if these settings are best for this source but parametlicly (and visually too) gived a better result.
As a conclusion , x264 > VP7 for SSIM test.
x264 < VP7 for OPSNR test.
But maybe there is oportunity to find better balance between OPSNR and SSIM changing values for RC.

You just saying that VP7 is better because it´s price to receive a beta from on2. It´s obvios. ;) :p

[Sagittaire]

In fact VP7 (commercial build with PIV) is always better than x264 (rev 158 or rev 174). Update for x264 between rev 158 and rev 178 are praticaly only speed optimisation ... I wait veritable update from rate distortion for x264 ...

[Koepi]

This thread leads to nothing but flames. Closed.