New xvid binary?

[gizmotech]

Hi guys,

Just a quick question, is there any chance of a new Xvid Binary? Our encoding group has been doing a lot of experiementation, and we've noticed a lot of random block errors from teh r-d Trellis option, we were hoping that a new binary might be available with newer r-d trellis code.

Not trying to push ya, just thought I would ask. Keep up the good work guys.

GizmoTech.

[superdump]

No work is being done on dev-api-3 anymore. Koepi's build is the last of that line I believe. However CruNcher has compiled the CVS of dev-api-3 and it does perform slightly better, so maybe, if you're very very lucky, Koepi might compile another dev-api-3 build.

[gldblade]

What's happening now with XviD anyway? I've read much of the posts on this board, and the XviD devel archives, but I still don't have much of a sense of what's currently going on.

So, where are we headed?

[amango]

@gldblade

Just don't use R-D, if it cause trouble for you. It's better not to use it. I made some encodings with it, and I had the same problems with those artefacts. R-D decreases filesize but on the other hand it also decreases quality. Especially for cartoons, R-D is not a good option.

[superdump]

amango: If you're talking about dev-api-3 then I can possibly agree with you because of the (iirc) inverted luma block bug in trellis. I haven't experienced this myself and also (iirc again) i think it only happens on dark sections, but I may be wrong on that.

Just recently through a couple of tests I found that using all the usual options, h.263 and trellis and NO B-Frames worked better on cartoons than when using B-Frames. Afaik trellis cannot work on B-Frames and it appears to have awesome power with cartoons. I found the full options encodes to be a touch less noisey and a little more blurry than the ones without B-Frames. It's up to you which you prefer.

As for where XviD is going: the non-magical v1.0 release will come out of dev-api-4 WHEN IT IS FINISHED. Development progress varies, but sit tight, it will come eventually.

[celtic_druid]

If you want a new binary, just download and compile the devapi4 source.

[Angrychair]

Originally posted by celtic_druid
If you want a new binary, just download and compile the devapi4 source.

Haven't had any luck doing that. Got some pointers for success? Always get the following error:

plugin_2pass2.c
plugin_dump.c
c:\cvs\xvidcore\src\plugins\plugin_2pass2.c(963) : error C2520: conversion from unsigned __int64 to double not implemented, use signed __int64
c:\cvs\xvidcore\src\plugins\plugin_2pass2.c(963) : error C2520: conversion from unsigned __int64 to double not implemented, use signed __int64
c:\cvs\xvidcore\src\plugins\plugin_2pass2.c(969) : error C2520: conversion from unsigned __int64 to double not implemented, use signed __int64
plugin_lumimasking.c
plugin_psnr.c
plugin_single.c
decoder.c
encoder.c
c:\cvs\xvidcore\src\encoder.c(1735) : warning C4018: '<=' : signed/unsigned mismatch
c:\cvs\xvidcore\src\encoder.c(1742) : warning C4018: '>=' : signed/unsigned mismatch
xvid.c
c:\cvs\xvidcore\src\encoder.c(721) : warning C4761: integral size mismatch in argument; conversion supplied
c:\cvs\xvidcore\src\encoder.c(721) : warning C4761: integral size mismatch in argument; conversion supplied
c:\cvs\xvidcore\src\encoder.c(721) : warning C4761: integral size mismatch in argument; conversion supplied
c:\cvs\xvidcore\src\encoder.c(726) : warning C4761: integral size mismatch in argument; conversion supplied
c:\cvs\xvidcore\src\encoder.c(726) : warning C4761: integral size mismatch in argument; conversion supplied
c:\cvs\xvidcore\src\encoder.c(726) : warning C4761: integral size mismatch in argument; conversion supplied
c:\cvs\xvidcore\src\encoder.c(731) : warning C4761: integral size mismatch in argument; conversion supplied
c:\cvs\xvidcore\src\encoder.c(731) : warning C4761: integral size mismatch in argument; conversion supplied
c:\cvs\xvidcore\src\encoder.c(731) : warning C4761: integral size mismatch in argument; conversion supplied
Error executing cl.exe.

libxvidcore.lib - 3 error(s), 11 warning(s)

If you're serious about compiling the the codec, I'd appreciate some pointers. If you're just suggesting that the bourgeois eat cake, then I apologize for posting in advance.

[BoNz1]

You don't have sp5 for vc++ installed you will need this as well as nasm to build it.

[Angrychair]

Originally posted by BoNz1
You don't have sp5 for vc++ installed you will need this as well as nasm to build it.

I do have SP5 installed. (rather, I ran the installer, though I've heard that there's some sort of problem updating Visual studio under windows2000. Any tips?)

[celtic_druid]

Try using the Intel compiler.

[Angrychair]

Originally posted by celtic_druid
Try using the Intel compiler.

Thanks for the suggestion, but I don't have the intel compiler. I'm more interested in getting service pack 5 properly installed, though. I can't be the only person here that uses windows2000 and is trying to update VS6, but it just doesn't seem to be working.

Ok, I got it working. I needed this: http://msdn.microsoft.com/vstudio/downloads/tools/ppack/download.aspx

It's the processor pack that apparently adds processor support that is needed in this case.

Sorry for the trouble, and I hope that what I've found there will help someone else.

As an aside, the new xvid dev-api-4 interface sure is different, can't wait to get cracking on it. =)

[int 21h]

You extracted the files for SP5, and then ran the installer right? Some people forget to run the installer...

[Angrychair]

Yeah, I did run it, but there's nothing in VC++ to indicate what SP you're using. However, I found the solution to the problem now, thanks for the help.

[superdump]

Angrychair: If you have any questions about the GUI (i.e. where something is) just give a shout on here and either me or someone will probably give a reasonably speedy reply. Although, having said that, I'm out all day today. :)

Anyway, the GUI may change as many people are not happy with it. I personally think it's OK if a little different and a bit clunky with so many popup windows. I've made a suggestion for an alternative but we'll have to see if anyone feels like coding that.

Have fun! :D

[RadicalEd]

ANGRYCHAIR, I LOVE YOU.

I've been having the same exact problem forever, and I've been waiting to hack cartoon mode into the vfw to test out. Thanks a ton.

Edit:
"ERROR: The Visual C++ 6.0 Processor Pack requires either VisualStudio 6.0 Enterprise or Professional."
I have Visual Studio 6 Enterprise.
;_; I give up.

[Angrychair]

Originally posted by RadicalEd
ANGRYCHAIR, I LOVE YOU.

I've been having the same exact problem forever, and I've been waiting to hack cartoon mode into the vfw to test out. Thanks a ton.

Edit:
"ERROR: The Visual C++ 6.0 Processor Pack requires either VisualStudio 6.0 Enterprise or Professional."
I have Visual Studio 6 Enterprise.
;_; I give up.


Hm. I also installed a couple of other patches before I installed the processor pack. It was the C++ library update (about 1.5 megs, don't remember the link, though, sorry)

I also updated mdac to 2.8, I believe, and rebooted a few times.

Strangely, though, I can't find cartoon mode listed anywhere in the options dialogues for the version I've compiled. It's definitely a much clunkier interface, though, as the options aren't all intuitively placed.

[superdump]

AngryChair: Cartoon mode isn't in the GUI yet.

All: Cartoon mode doesn't provide much, if any benefit, from my testing. It's something very simple in the code but it didn't appear to do much. So don't get all excited about it.

[Angrychair]

Originally posted by superdump
AngryChair: Cartoon mode isn't in the GUI yet.

All: Cartoon mode doesn't provide much, if any benefit, from my testing. It's something very simple in the code but it didn't appear to do much. So don't get all excited about it.

Hm, how can one enable it, then? I'm rather disappointed to see alt. curve gone, I found it very useful for the fact that you could use relative quality to automatically keep your quantizers in line on an encode rather than trying to play the capping game.

The GUI definitely needs some simplification, though. There are what? 4 or 5 seperate dialogue boxes? The b-frame threshold setting isn't even in the same dialogue box as the b-frame settings.

Also, any insight on what the reduced resolution option is? I believe it was causing my encode to crash when I tried to enable it.

[bond]

download the latest official xvid version 0.9.2 (stable) here (http://ciberia.ya.com/kalamanweb/dvdadivx/xvid092.zip)! it includes all advanced features like qpel, b-frames, vhq, etc...
(consider koepis stable build as outdated)

to get a small look inside DevAPI4, download this (http://ciberia.ya.com/kalamanweb/dvdadivx/XviD-DevAPI4.zip) build!
no dont never use this build for encoding!!!

[AmiRage]

Originally posted by bond
download the latest official xvid version 0.9.2 (stable) here (http://ciberia.ya.com/kalamanweb/dvdadivx/xvid092.zip)! it includes all advanced features like qpel, b-frames, vhq, etc...
(consider koepis stable build as outdated)

to get a small look inside DevAPI4, download this (http://ciberia.ya.com/kalamanweb/dvdadivx/XviD-DevAPI4.zip) build!
no dont never use this build for encoding!!!
Thanks, but all I get is ...

La página a la que estás intentado acceder NO EXISTE o bien NO ESTA DISPONIBLE.
Por favor, comprueba si la escribiste correctamente.

... when trying to download these files.

[bond]

try the first two links on http://www.xvid.net

[Lobuz]

At canalxvid.com (http://www.canalxvid.com) at forum is DevApi4 build from 10-08-2003

Regards
Lobuz

[amango]

@Koepi, Nic, Umaniacs

Can you make a new build out of the latest xvid 0.92?

[HarryM]

Originally posted by Lobuz
At canalxvid.com (http://www.canalxvid.com) at forum is DevApi4 build from 10-08-2003

Regards
Lobuz

I test this 'buggy syskin's' devapi4 build. It has suspectly very high compressibility(?) at using b-frames compare to the last Koepi's 24-06-2003 build....

E.g. (same setup; b-frames=3; q-pel, VHQ=1, chromaME, quant=2)

devapi4 syskin 10802003 --- 12 042 240 bytes
devapi3 Koepi 24062003 --- 18 180 096 bytes

???

[sysKin]

OK guys, you b0rked something. XviD 0.9.x series does NOT have all the things you talk about. The newest of them is 0.9.3 and even if this version is new, the code behind it is one year old.
It does *NOT* have qpel, b-frames, vhq, etc.

Tell me where you have it from, or complain directly to the person who said so on his webpage.

superdump wrote
Afaik trellis cannot work on B-Frames (...)Yes it does, no problems there. I might have forgotten about putting this part to dev-api-3 for Koepi so maybe b-frames are intact in it, but it won't change much anyway. Definitely works in dev-api-4.

[bond]

syskin,
i found the 0.9.2 (thats the last version shown on xvid.org) build on canalxvid (at least they claim that it is build from 0.9.2 source) and it includes qpel, etc...

[Kurosu]

Originally posted by HarryM

devapi4 syskin 10802003 --- 12 042 240 bytes
devapi3 Koepi 24062003 --- 18 180 096 bytes

???
Filesize alone means nothing, even with apparently "same" settings. In that very case, it's not even worth comparing, as the filesizes show something is borked. Otherwise, you'd need some quality metric such as PSNR to evaluate coding performance of the 2 builds.

@Angrychair
Compile is fine here with VC++ enterprise, latest Service Pack and Processor Pack 5 on a W2K SP4 box. But it also has CoreSDK and DX8.1 SDK installed (your errors aren't related to that, though - just do a cvs update).

@Anyone interested
1)
(that may get me flamed)
cartoon mode should be enforced (ie would always be on) by the following lame hack:
in vfw/src/codec.c, inside LRESULT compress(CODEC * codec, ICCOMPRESS * icc), once the frame struct has been allocated, you can add frame.vop_flags |= XVID_VOP_CARTOON;

2)
This mode, from my limited understanding, helps to choose more (0,0) vectors. In anime, SAD between adjacent blocks is very similar and could lead to false motion detection (moving blocks)
Then, the threshold under which AC coefficients are discarded (because of the TOO_SMALL_LIMIT) is increased by a factor 3 in order to keep more coefficients (and then avoid blocking artifacts on low detailed surfaces).

(If a XVID developper doesn't like the later part of this post - and there are reasons I'd perfectly agree to -, please message me so that I edit it)

[AmiRage]

Originally posted by sysKin
About the CanalXVID's build: it's my build, just go on IRC (my signature) and you can get it directly from me. I'll update it with newest bugfixes in an hour.
Could someone "publish" this new build from sysKin outside IRC? Thanks in advance.

[Angrychair]

Oh yeah, for anyone that still may need it, this is where I got that libraries update:

http://www.microsoft.com/windows98/downloads/contents/WURecommended/S_WUFeatured/Libraries/Default.asp

Good luck, testing out that cartoon mode thing now.

[Mashabak]

Hello, this post affected me, because I was who compiled that official version (0.9.2).
First of all, I have to say I am sorry for this incident, because it is an important error, since the core of the codec (libxvidcore.lib) really belongs to the source core of 0.9.2 version downloadable from xvid.org, and that my error was in that I used a proyect of VFW(Video for Windows) too much new (latest one of CVS) for the Windows compilation, because this was the *only* proyect available on this web to be downloaded.
I am sorry for all the problem it has been caused, and I must add that I have recompiled the version using a previous proyect of VFW , and now, the codec only has the options that really must have.
Also say that devapi4 compilations in the web are ONLY for testers and that is write on principal post about these versions, and all versions that we bistribute are only for people that can´t compile the codec itself and want to test it.

Greetings and sorry for my poor english, Mashabak.