New Xvid FAQ...feedback requested.

[crusty]

Hi all!

Since Doom9's FAQ here is rather outdated and the new xvid builds changed a lot of things (not only looks, but also about what is safe/MPEG-4 compliant/recommended etc.) I decided to write a new FAQ for Xvid. I started out with rewriting Doom9's FAQ, but I ended up doing a total rewrite. It's turned out to be quite big and elaborate and I hope you like it.

Basic setup currently is in 5 main chapters:
Part A: Introduction
Part B: Getting it working
Part C: Options, Schmoptions....
Part D: Troubleshooting
Part E: Feedback and questions
Appendix (in which stuff like links, downloads, faq contributors etc.)


You can find it at:
http://www.vslcatena.nl/~ronald/docs/xvidfaq.html
(the site is a bit slow now and then)

What I really like is some good feedback and discussion about it.
In descending order of importance I would like feedback about:
-Content errors
-inconsistencies
-useful additions
-Spelling errors
-makeup errors/html tips
-anything else I forgot

I especially like some good tips/explanations on the following stuff which I haven't yet added or are considered incomplete:
B6. I want to compile Xvid myself. How do I do that?
C7: Single Pass 'More' settings explained
C9. Two-pass second pass 'more' settings explained

Kudos go to Koepi, matmike, Nic, sysKin, Snowbeach for helping me out on taking out many errors, inconsistencies and other crap :D.

Enjoy!
Crusty

Edit: Small update to FAQ:
Moved C15 to C16
New C15: "Things that work but are not MPEG-4 compliant."

[Soulhunter]

DAMN NICE FAQ... !!!

[kempodragon]

Very well organized format, I'll have to do a full review this weekend.

[Koepi]

Hehe, I put the link to your FAQ in the forum XviD FAQ (I did the mistake and took a short look what's currently in our FAQ. Amazing. I thought my last changes were some weeks ago. Damn. How fast weeks add up to months ;) ).

I still have some points to add (don't think i'm a jerk, I just want to help):

- most distracting from the content is for me to read "usefull" - it's "useful", with 1 "l". I know. That's picky from me.

- B2: my builds now don't even need a mmx capable CPU, I'm quite sure. though I wouldn't understand why anyone would use a 486 for encoding/decoding.

- B4: Gamr's builds are also insta-builds (=automatically compiled if CVS changes are detected, checking interval is the same is uManiacs, 20 minutes if I'm not mistaken). With all the consequences you mentioned for uManiac ;)

- D1: Loading defaults isn't mandatory after installation of newer (means: 1.0-xxx-...) builds. The registry settings will get dleted upon installation, and thus with the first codec usage the defaults get loaded anyways :) But - if you run into troubles, "load defaults" and doing the coec setup again might help as well.

Well, those are some points which i've noticed in my current, fast review (no time as usual :( ). I hope this helps, and keep up the good work!

Thanks,

regards
Koepi

[crusty]

Roger, will update asap.

Any other non-MPEG-4 compliant stuff besides what I put in C15 ?

EDIT:
Replaced al 'usefull' with 'useful'
B2: removed mmx requiremnet

B4: Added GamR and mention about CVS-tree
D1: Altered it a bit, check it now.

@ RadicalEd below:
If WMV(9) isn't MPEG-4, what is it then?

[RadicalEd]

WMV 9 was incorrectly listed as an mpeg 4 codec.

[Leak]

Originally posted by Koepi
- B4: Gamr's builds are also insta-builds (=automatically compiled if CVS changes are detected, checking interval is the same is uManiacs, 20 minutes if I'm not mistaken). With all the consequences you mentioned for uManiac ;)

- D1: Loading defaults isn't mandatory after installation of newer (means: 1.0-xxx-...) builds. The registry settings will get dleted upon installation, and thus with the first codec usage the defaults get loaded anyways :) But - if you run into troubles, "load defaults" and doing the coec setup again might help as well.

Since Gamr got his autobuilder going again, I just gave his latest build a try and noticed 2 things:

- He's still using an NSIS installer that won't automatically detect and uninstall your RC2 build, which might lead to trouble - I guess mentioning this in the FAQ wouldn't be a bad idea

- His build also didn't reset the registry settings, so mentioning hitting "Load Default" couldn't hurt either

Just my .02 EUR...

np: Markus Guentner - Lost Paradise (Audio Island)

[Koepi]

I meant to send him my updated NSIS script - but before i could do that, ReactOS took over my partition table :( It'll be some major work to get innosetup/istool running on his WINE, but if he's online again (i haven't seen him for some weeks online) i'll send him the _new_ installer system script.

But well, if you know the restrictions which apply support-wise to insta-builds, is it really necessary to mention? Someone not familiar with xvid shouldn't even _think_ about an instabuild.

Just my 2 bucks :P

Regards
Koepi

P.S.: hey crusty, does "schmoptions" imply anything in dutch? Because I don't get the joke :-/ Maybe i'm too much technical orientated though. (Btw., the thing about that "usefull"-stuff is just meant friendly. I got some very unfriendly emails when i wrote usefull in the ancient XviD Options Explained... I want you to be safe from such feedback.)

[Leak]

Originally posted by Koepi
P.S.: hey crusty, does "schmoptions" imply anything in dutch? Because I don't get the joke :-/ Maybe i'm too much technical orientated though.

Sounds like "Consequences, schmonsequences" to me, which in English is often used to downplay possible consequences of someone's actions... :)

And while I'm reading the FAQ:

- A17's "XviD will, however, incorporate new features in the future such as B-frames" sounds a bit dated to me, as it's been supporting B-frames for ages... :)

- C4 should probably note that the tab not only is just informative, but that XviD doesn't take anything but allowed options and maximum frame size into account (i.e. it doesn't limit the bitrate as the information might suggest); yeah, the dialog says so, but sometimes it helps more really driving the point home... ;)

- C8 could use a mention that by turning off fast first pass you get a perfectly valid video file in the first pass

np: Plaid - Kortisin (Not For Threes)

[KpeX]

Originally posted by crusty
If WMV(9) isn't MPEG-4, what is it then? It's proprietary. But I believe the spec is now somewhat open. More info from the search button ;)

[crusty]

- He's still using an NSIS installer that won't automatically detect and uninstall your RC2 build, which might lead to trouble - I guess mentioning this in the FAQ wouldn't be a bad idea
- His build also didn't reset the registry settings, so mentioning hitting "Load Default" couldn't hurt either

Roger, I will add something along those lines in both B4 and D1
P.S.: hey crusty, does "schmoptions" imply anything in dutch? Because I don't get the joke :-/ Maybe i'm too much technical orientated though. (Btw., the thing about that "usefull"-stuff is just meant friendly.
It doesn't imply anything other than have a bit of fun when reading :)
It's something like what Leak says. And np about the spelling error..I'm no English native either so I welcome any spelling corrections as well.

@A17: Well there you have one of the few lines of Doom9's FAQ that are still present in my FAQ...and it's outdated. LOL
Will remove asap.

@C4 I thought I already put it in..must be somewhere else where I added it. I will put another warning in C4 just to be sure. In bold. :D

C8: Yeah that button is extra in RC2 and I didn't update that part yet..I was aware of it tho.
Good that you mention that it is a valid file, I will add that.

About WMV9...I thought it was a descendant of the MPEG-4 codec of MS that later got hacked and became the DivX 3 codec. Isn't it so?


EDIT: OK, updated A17, B4, C4, C8 and D1
Check them now and see if you like the results :D

[iago]

Very nice indeed! Good work, crusty! :)

[mikeX]

hi there, just caught this nice thread and started reading the faq, it's very late here though and i'm sleepy, so i only took a quick glance at it (and got up to A4)

everything i say is to be taken as suggestions in good faith, not nitpicking or complaining :)

-inconsistencies: xvid and GNU/Linux!
(i know the average guy who stamble's into the faq uses window$, but that doesn't mean everyone using linux is an xvid/mpeg-4/video encoding/decoding guru... - at least clarify that you are talking about xvid for window$ from the beginning) (also sorry if i missed anything, i just hit find 'linux/unix' and came up with nothing, will read the whole thing tomorrow)
(i'm willing to try and figure out some very basic q&a if you want)

-Spelling (or syntax) errors:

Foreword
"explanations of Xvid options tended to be rather short and cryptic (if there was one at all)"
--> if there were any at all

A1
"GNU Public License"
--> GNU General Public License

A2
"( or more accurately to program chain that contains the codec)"
--> (or more accurately to the/a program chain that contains the codec)

"-Then the codec does it decompressing "
--> -Then the codec does it's decompressing

keep at it man ;)

[RadicalEd]

Originally posted by crusty

About WMV9...I thought it was a descendant of the MPEG-4 codec of MS that later got hacked and became the DivX 3 codec. Isn't it so?

The sources are completely proprietary as KpeX said, there's really no way to tell exactly what's in it. Microsoft has denied a relationship between MPEG 4 and WMV 9, though, and it certainly isn't compliant with any MPEG 4 decoder, so that's as close as we're going to get. If you remember back in the days of WMV 7 and 8, an ISO MPEG 4 codec and msmpeg4v3 were included seperate from WMV in WME, implying that WMV has been a seperate project from the start.

[bond]

Originally posted by crusty
the MPEG-4 codec of MS that later got hacked and became the DivX 3 codec.altough the name says something else ms mpeg-4/divx3 are not mpeg-4 compliant!

[crusty]

mikeX: Roger, spelling errors updated. Thanx. Keep up the nitpicking! :D

As you say, it's indeed mostly windows oriented.
I know linux a bit, but mostly as a server OS. I don't use it as a Desktop OS.

I'd be more than happy to put some linux stuff in if somebody tells me some details about it. (Same goes for Mac/BeOS/Commodore 64/whatever :) )

That goes basically for anyone: If you've got a better written explanation for a question/option, I'd be more than happy to replace mine with it. I just want it to be as good as possible.

-Removed wmv-9 as an MPEG-4 codec.
BTW: How about VP5/6, is that MPEG-4 or not?

[virus]

Originally posted by crusty
BTW: How about VP5/6, is that MPEG-4 or not?

No it's not. Purely proprietary format.

[Soulhunter]

Maybe something like this could be added ???

-> I get pixolous distorsions in red arears !!! (http://forum.doom9.org/showthread.php?s=&postid=390365)

Bye

[mikeX]

well i have to say that this faq is WAY MORE than just an xvid faq, but one settles for more a lot easier than less ;)

some more (a lot really) 'nitpicking' :D (got up to C6)

spelling/syntax:

A2:
Typically, a codec codes only for one type of data it's specificly designed for.
('the' type of data sounds a bit better)

A3:
In software usually means in some form of codec or decoder software,
(hmmm, maybe 'encoding or decoding'?)

more than 1 B-frame, which Xvid can handle but Divx cannot.
(well it can with the new workaround in rc2, or do you mean encoding wise?)

'that can also play divx' vs 'besides just Divx.'
(capital D, you could run a X/xvid/D, d/Divx/X find&replace with XviD, DivX)

A4:

(well, atleast in theory).
(at least)

Compression FAQ @ www.faqs.org
(maybe hyperlink the whole thing)

the car(s) dead center in his view.
(maybe just 'centered')

video information consists of this type of situation.
(maybe 'falls in that case')

The codec divides the picture up in 8x8 blocks
(maybe 'divides into'?)

DCT takes the detail in those blocks and performs a technique called the quantization process on it. The 8x8 blocks then no longer contain pixels but frequency values.
(i'm confused, are they called frequencies after the DCT or after the quantization process?)

THere are other types

look at a Keyframes without

P-frames only contain the difference between it and the previous frame.
P-frames only contain the difference between them and a previous frame.
or A P-frame only contains the difference from a previous frame.

and in this frame
(and in the current frame?)

that's alot -- the sentence doesn't really need to be in ()

-B-frames are Bi-directional
(maybe 'bi-directionaly predicted'?)

It tries to reference the previous frame, the next frame, or a mix between the two, and choses the way that works best.
(what does? the codec or the frame?)

A5:
occasional thankyou.

A7:
then come over back here
(then come back over here)

A10:
Don't Panic.

A12:
Basically two-fold:
Basically twofold:

without giving a error during compiling.
without giving an error during compilation.

A14:
but have less features or slower.
but have less features or are slower.

B2:
a computer with a 32-bits cpu.
a computer with a 32-bit cpu. (don't know if it's right but it sound's better, doesn's it?)

B3:
extra component requieres at
extra component requires at

note:: you might also wanna mention built-into-motherboard sound/network chipsets(/internal modems) which rely mostly on the cpu to do their processing

C1:
all the options are what makes
(isn't the syntax wrong, can't figure out the correct one...)

playable with other codecs than Xvid,
playable with codecs/decoders other than Xvid,

crawl but will look stunning.
(maybe 'crawl but give stunning results.'?)

containing....more
containing.... more

(in this example virtualdubmod,
you said 'virtualdub' in the beginning, might wanna replace with virtualdub(mod) for the whole faq to be sure...

C3:
that will allow give you unrestricted access

C3b:
is a first 'psychological' innovation in XviD; it is supposed make use of the
is 'psychovisual?' innovation in XviD; it is supposed to make use of the

C3c:
Especially most interlaced DVD's and VHS video's are
(wrong syntax/spelling i think)

C3e:
These macroblocks motion vector
These macroblocks' motion vector

C3f:
vector and whole picture moves.
vector and the whole picture moves.

C3h:
bi-directionally
bi-directionaly

C6:

uses the compressability data
uses the compressibility data
(dictionary says the right adj for compress is 'compressible')

encoding a life feed,
encoding a live feed,

notes/suggestions:

B6: compiling xvid:

some early startpoints:
for windows: proprietary compiler: MSvisual studio, Inter compiler
free/gnu: mingw(?), cygwin(?), bloodshed(?) (dshow problems?)
for linux: piece of cake...

some links: http://www.bloodshed.net http://www.mingw.org http://www.cygwin.com


C1: what can you tell me...

Might wanna mention somewhere that you are specifically talking about xvid >= rc1 gui
(e.g. 'More' is '...' in the betas, not to mention the 0.9x tree gui...)


C9:

Simple 'Overflow Payback Example" (all 3 options) from personal experience:
Default setting of 5 will give a higher quantizer for the majority of frames as to even out the main quantizer distribution to a small zone.
A setting of 80 is likely to spread the quantizer distribution to a wider zone (could range to the whole zone of 1-31, depends on the source/desired compression), giving many frames a lower quantizer than before but also giving a higher quantizer to (probably) equally more frames.
So what does that lead to? one frame can have a really low quant (and look really good, full of detail) while the following frame has a really high quant (looking all messed up and blocky) giving a bad overall impression...

i'm gonna try and start writing down some linux stuff (i really suck at writing stuff...)
a starting point for some linux q&a could be this thread (http://forum.doom9.org/showthread.php?threadid=67976&highlight=settings+AND+linux)

[yaz]

Originally posted by Leak
... Gamr ... his latest build ...
- He's still using an NSIS installer that won't automatically detect and uninstall your RC2 build, which might lead to trouble - I guess mentioning this in the FAQ wouldn't be a bad idea
- His build also didn't reset the registry settings, so mentioning hitting "Load Default" couldn't hurt either
... &
- it delivers a dsf seems to be quite outdated

faq is prettypretty cool ! thx a lot !!

the best
y

[crusty]

@MikeX:
Damn that's a big list!
Thanks for the continued nitpicking. :)

A2:'the' type of data sounds a bit better
If you look again, you will see that it's a contrast between interoperable and non-interoperable codecs. That's why I chose 'one' type instead of 'the' type. Sounds logical?

A3:
How about this:
"Typically MPEG-4 applications are implemented either in software or in hardware. In software usually means in some form of encoder or decoder program, while hardware implementations are typically electronic chips or circuitry designed to encode and/or decode MPEG-4."

well it can with the new workaround in rc2, or do you mean encoding wise?)
Well the workaround is so fresh you can still taste the programmers' coffee in it, so I'll wait a bit for more feedback before I change it.
And of course DivX doesn't encode more than 1 B-frame.
Another typical example would be Xvid's GMC.
Added:
"Another example would be Xvid's 3-warppoint GMC."
capital D, you could run a X/xvid/D, d/Divx/X find&replace with XviD, DivX
That's indeed one of those document-wide inconsistencies that I still have to handle.
Replaced.

A4:
maybe hyperlink the whole thing
I did, didn't you check your bottom bar @ your lower left when you highlighted it?
maybe just 'centered')
That's a nice tip. Replaced.
maybe 'falls in that case'
Nope, even better:
Replaced:
"information consists of this type of situation"
with:
"footage falls in this category"
Also check the added italics in the next sentence.
(i'm confused, are they called frequencies after the DCT or after the quantization process?)

I'm no expert, but I think the DCT transforms the texture information into the 8x8 block 'mean value/frequencies'-combo, and then the quantization process drops the frequencies.
The motion search process works mostly with the macroblocks, but if I understand correctly higher motion precisions also work on lower scale, including 8x8 blocks all the way up to subpixel-precision.

THere are other types + look at a Keyframes without
Typos...removed
or A P-frame only contains the difference from a previous frame.
Altered to:
"P-frames only contain the <i>difference</I> from the previous frame."
and in the current frame?)
Altered to:
"and in the current frame it went from red to green, then this frame would only contain the altered information"
Better?
>More typos altered...<
maybe 'bi-directionaly predicted'?)
Sounds good...replaced.
what does? the codec or the frame?)
That needs some lumination indeed...how about this:
"The codec tries to reference the previous frame, the next frame, or a mix between the two, and choses the way that works best. If the frame references only the previous frame it becomes a P-frame, otherwise it becomes a B-frame."
(Please anyone, tell me if this is technically incorrect.)

A5:
Altered to 'thank you'

a7:
Woot..! :eek:
just removed the 'over'

A12:
OK, altered.

A14:
"but may have less features or are slower."

B2:
typo..altered

B3: idem
Added: "(including 'on-board' ones)"

C1:
Added: "(Note: 'More' is '...' in the betas)"
Altered the "playable with other codecs than Xvid" sentence a bit.
Check it out.
Altered to: "crawl but will give really great results"

Will make Vdub(mod) consistent later on..thanks for mentioning it.

C2: removed 'allow'

C3b:
Altered

C3c:
Think this:
"Especially, most interlaced DVD's and VHS video's are.."
Then it will make more sense. I don't put the comma in because it's only meant to be 'thought' there, not to 'be' there.

C3e:
Altered.
Also altered any 'vector' to 'motion vector' and altered the value explanation to:
"(as a sort of two-dimensional X,Y value)"
Which is much more correct.

C3f:
Altered the whole makeup of the answer a bit, mainly for readability.

C3h:
Check: http://dictionary.reference.com/search?r=2&q=ly
It's directional + ly = directionally

C6:
Altered.

Notes/suggestions:
B6:
Just quickly added the links you gave.

I think I will add a B8:
'How do I use XviD under Linux?'

Pff... a bit tired now...gonna chill a bit.

Keep the suggestions coming!

[mikeX]

just glad to be helping out anyway i can ;)

Regarding Linux integration into the faq, i agree one or two questions are enough, since it's too much trouble integrating it into the whole thing (program chains etc)

it could be something like:
:scared: HELP, what are all those windows programs you keep mentioning, my pc says GNU/Linux at startup (just kidding)

seriously now, it could be:
Q: I see this faq concentrates mostly on using XviD under Windows, but i've heard you can use it on any Unix-like platform, like GNU/Linux. So what's the status on xvid support there?

A: Xvid can run on any unix-like platform, such as ... ...
Most of what's said in A2 applies for such platforms as well...blah blah blah...
the most commonly used applications for encoding with Xvid under GNU/Linux are 'transcode' and 'mencoder' (links etc) which come with the proper communication module for the codec (be sure to get an up to date one for the latest versions (>beta) from http://ed.gomez.free.fr or from transcode's or mencoder's cvs)
these are command-line applications but have many frontends providing them with a GUI (some common ones dvd::rip,...)
moreover xvid configuration is not handled the same way as in windows, so don't look for the configuration GUI, it's not there.
configuration is usually handled from the encoding application with the help of configuration files.
The same options apply offcource but whether or not you have access to them depends mostly on the application you use at the moment.
Check out this thread for more on that:
http://forum.doom9.org/showthread.php?threadid=67976


what do you think (these are just some basic points)?
i'm gonna post back with a more detailed version

regarding B6:
it would be great if a dev could answer on the state of the mentioned compilers (no details, just what works and what doesn't with each one)
the 'INSTALL' file that comes with the xvid source code mentions only MS Visual Studio for windows compilation...
btw that file has some useful info on compiling xvid, it's under '/xvidcore-x.x/doc/INSTALL' in the source package (at least in the bz2 archive i have)

here are some useful threads i dig up from the xvid.org forum:
ms visual c (http://www.xvid.org/modules.php?op=modload&name=phpBB2&file=viewtopic&t=1202)
intel compiler (http://www.xvid.org/modules.php?op=modload&name=phpBB2&file=viewtopic&t=549)
a guide for compiling xvid under MS VC (http://www.discdude.net/xvid/compile.html) (taken from the first thread)



back to nitpicking, :D

A2:
yeah i see what you mean, it just didn't sound right, it's no big deal

A4:
Compression FAQ @ www.faqs.org:
yeah i saw it, i actually meant something like this:
Compression FAQ @ www_faqs_org (http://www.faqs.org/faqs/compression-faq/part2/section-2.html):
instead of
Compression FAQ @ www.faqs.org (http://www.faqs.org/faqs/compression-faq/part2/section-2.html)

gets converted from whatever it was to a colour space called YUV2.
shouldn't that be YV12???

I'm no expert, but I think the DCT transforms the texture information into the 8x8 block 'mean value/frequencies'-combo, and then the quantization process drops the frequencies.
The motion search process works mostly with the macroblocks, but if I understand correctly higher motion precisions also work on lower scale, including 8x8 blocks all the way up to subpixel-precision.
hmmm, i'm still a bit puzzled, i'll have to check the 'understanding a QM' thread again :)

A7:
There also newer and more flexible file formats
(There are also newer and more flexible file formats)

A good place to start would be by reading the Avi overview on this site or the faq here.
(maybe 'A good start would be to read the Avi overview on this site or the faq here.')
(general note:I think ogm can be also referensed as 'Ogg Media File'(not sure though))

A14:
"but may have less features or are slower."
(hmmm, come to think of it be slower fits better, right?)

B3:
Added: "(including 'on-board' ones)"
(i would say especially on-board ones) ;)

C1:
maybe also add something like:"Expect different look/options/terms for versions < xvid 1 beta"

C3c:
hmmm, still, i think video doesn't have a plural

Sorry about that 'bi-directionaly' stuff, i was in a bit of a 'brain-dead' state (still am most of the time) and couldn't really think... you 'll have to change it to ll on my 'bi-directionaly predicted' suggestion at A4

nice fixes/alterations overall ;)

- some additions -

C10:
imput field and slider are inactive.
input field and slider are inactive.

C11:
Sensitivity will allow you tweak the amount
Sensitivity will allow you to tweak the amount

the amount of B-frames; A negative
'the amount of B-frames; a negative' or 'the amount of B-frames. A negative'

along with that final warning on zones you could also mention there is a known problem with weight zones < 0.2 (at least atm)

ps the italics seem like a good idea :)

[sysKin]

Originally posted by mikeX
the 'INSTALL' file that comes with the xvid source code mentions only MS Visual Studio for windows compilation...No it doesn't - point 1d explains how to compile win32 build on any unix (including mingw and cygwin, but also anything else, like linux).

Also let me point out that point 2 explains MS Visual Studio 6. VS .NET works but is not supported, mostly because it makes errors (big ones) converting project files. Other compilers are just not supported.

:)Radek

[mikeX]

No it doesn't - point 1d explains...
sorry, i meant compiling in windows, not for windows in general (cross compiling)

:)

edit: how about bloodshed, etc... any known errors there?

[sysKin]

Originally posted by mikeX
sorry, i meant compiling in windows, not for windows in general (cross compiling)Um, cygwin and mingw are windows programs :)

[mikeX]

well, i thought mingw was also needed to cross compile from Linux :rolleyes: and that this was the case mentioned in the INSTALL file

so if what is said in the INSTALL file is correct, then you can't compile the decoder with such a program (mingw or cygwin), even from windows?

[atreya2011]

Nice FAQ. Can you modify section C9. "Two pass 2nd pass more settings explained" and explain in detail about "max overflow improvement" and "max overflow degradation". I want to know what they are meant for. Interested to learn.

:thanks:

Also there is a typo in Section "C3a. What is Quantization Matrix?"

Too put it short: It's the lossy filter by which you achieve higher compressibility by losing detail. The amount of detail lost is determined by the values of each individual quantum.

The first word...(To put it short)

[crusty]

gets converted from whatever it was to a colour space called YUV2.
To be honest...don't know for sure really...I have to do some forum digging to get that right probably.

A7:
Altered

A14:
I'm not sure...both sound a bit icky to me. I'll probably rewrite it in a later stage.

B3:
I know, but I already use Especially a few words before that.
Wouldn't be right style.

C1:
Added:
"<FONT SIZE="-1">(The following description is for the Vfw(Video For Windows)-interface for the RC2-build. In other builds the interface could be different, but you will find many if not all options present in one way or another)</font>"

C3:
Well, www.dictionary.com disagrees with you. :)
video - videos

C10:
Ok, did a document-wide 'replace imput with input'

C11:
Roger, altered.

Did some minor alterations to many parts. At the bottom of the document you will find a changelog and I've added the Timezone to last-update-time

I've added B8 and did some major alterations to B6 and B7, check it out. I Hope I got the linux part right. I couldn't find a proper home page for mencoder.

I would like some more feedback on C15: What's not MPEG-4 compliant

and I think I'm gonna add another question (probably in place of C15, moving it up)
Q: What are the current recommended settings?

Also: at
http://www.xvid.org/modules.php?op=modload&name=Sections&file=index&req=viewarticle&artid=3&page=1

I find the following Xvid flags:
"- XVID_ADAPTIVEQUANT : informs xvid to perform an adaptative
quantization.

- XVID_LUMIMASKING : infroms xvid to use a lumimasking algorithm. "

So what's the beef? I was under the impression that Adaptive Quantization was the new name for Lumimasking, but here it looks like they are two different options...
Is the Lumimasking flag there for compatibility reasons, or is it obsolete?

Keep it coming keep it coming..I've got plenty of keyboards here to trash. :D

EDIT:
sorry atreya2011, will update that error asap, I'm not very multi-threading ;)

[crusty]

OK, corrected

[crusty]

Working on C9 right now...but you have to have some patience....I'm figuring it out right now. :)

Any other comments?

EDIT:
Oh yeah, I figured a glossary would probably come in handy too...
But first I'll link to Doom9's glossary for the time being.

[crusty]

For people wanting to mirror my FAQ or covert it to Pdf, Doc, Sgi etc., please follow the following guidelines:

-always include at least my nick and the date-of-creation in the name of the document, using DD-MM-YYYY notation.
(Like for instance: XvidFAQ-Crusty-14-02-2004.pdf for today)
-always link to the current official location of the FAQ, I will add that info today in the top of the page. The newest version will be there. Current official location is at :
"http://www.vslcatena.nl/~ronald/docs/xvidfaq.html"
-Directly linking from websites to subparts of the FAQ, like linking to "http://www.vslcatena.nl/~ronald/docs/xvidfaq.html#B8", or to a download link, is allowed, but don't be surprised if I change any of them, as it's currently still in development.
-minor alterations in style are allowed when converting, changes in actual content will have to go through me first.
-Any mirror that wants to, will get listed officially inside the FAQ, (probably in the Appendix)

Thank you.

[crusty]

Current additions to C9 in progress:


Overflow treatment
'Overflow treatment' is the technique used to obtain a properly sized end result. Usually you specify a target filesize and the codec can either overshoot it's target, creating a file too big, or it can undershoot it's target, creating a file too small. Too counter this, overflow treatment can either allocate more bits than abolutely necessary, increasing filesize, or allocating less bits than really necessary, decreasing filesize. Obviously, the second process involves compromising quality.


Curve compression:
Normally the internal curve adjustment values (determined by the XviD developers after much feedback from users) are capable of delivering very nice results (I should say 'excellent' really), but if for one reason or another you want to, you can use these values to adjust the lows and highs of the bit allocation.
If you make a mental image of the curve allocation, you see a graph
with 'highs' and 'lows', sorta like hilltops and valleys. The hilltops are scenes with high bitrates and the valleys are scenes with low bitrates.
-The 'High bitrate scenes %' setting will take bits away from high bitrate scenes and give them back to the bit reservoir. (Think of a bit reservoir as a bucket full of bits wherefrom the codec hands out to each frame) Ergo, it will lower the hilltops, and the bits gained by this will be divided equally accross the entire landscape. This is useful if you really need to keep your encode within certain maximum parameters, like the maxima for a specific profile@level setting.
-The 'Low bitrate scenes %' setting will give extra bits to the low bitrate scenes, sort-of-like filling the valleys with sediment. The bits have to come from somewhere, so the codec takes all the frames in the entire encode and scrapes a few bits of them all.

So, basically, each setting favors compression of one of two possible extremes towards the average of the entire encode. The first takes down the hilltops, the second fills the valleys. The bits lost or gained are allocated accordingly to average out the entire clip.

EDIT:
Updated FAQ, now includes completed C9. Check it out.

[sh03z]

niiice job

I was looking for a good reading on XviD and finally here is one...


many thanks,

[ssjkakaroto]

excellent faq crusty! great job! :D

there are 3 lower cases on the topic links:
A2. what's a codec?
E4. how can I contribute to the XviD project?
c. XviD Developers

[george_zhu]

Very nice faq. Comprehensive.

One suggestion:

Why don't you add some explaination about "status windows"?

[atreya2011]

Obviously, experimenting with these settings may break filesize prediction completely as you're altering the basic settings of the underlying proces

Typo "proces" to "process"


'Cartoon mode' enables some mechanism in the motion estimation which drops (instead of encoding them) more macroblocks. The result is a more stable, a little less detailed image. Exactly what you need for cartoon like futurama or simpsons.


A More detailed explanation would be nice


:D

[crusty]

there are 3 lower cases on the topic links: Typo "proces" to "process"
Roger, typos corrected.
One suggestion:
Why don't you add some explaination about "status windows"?

Will do, as soon as I get around to some actual encoding again...LOL :D
<Intermezzo>No really, I'm in the process of cleaning up my pc before a reinstall of w2k. I always do a sort of 'controlled uninstall' previous to a reinstall of windows...that way I know exactly what programs to reinstall (must be 100+ at least, probably close to 200).

I b0rked my windows install because of testing too much software out for a network administration unattended install project</Intermezzo>
As soon as I have reinstalled my computer I will go back to some serious encoding.
A More detailed explanation would be nice
I took over that explanation from either Koepi or syskin. I will elaborate or replace that explanation in the near future asap.

I've also decided to underline all options, as that makes it easier to find them in the word-jungle that I call my FAQ right now. :)

[atreya2011]

First PDF Compilation.

Crusty's Unofficial XviD FAQ updated as of 15-02-2004 (http://www.geocities.com/atreya2011/index.htm)

Have Fun... :D

[virus]

just some suggested updates & mispellings detected (checked with my English dictionary)
(that's in sync with version Sunday February 15, 06.35 (gmt+1))

*A13 "Of course", not "Offcourse"
*A17 "compatibility", not "compatability"
*B2 you should also mention that cpu requirement depends on the frame size... certainly 320x240 is not the same as 1280x720 ;)
*B5 links to doom9.org's guides don't work anymore
*quote from C3:
"Step sizes will be small in the upper left (low frequencies), and large in the upper right (high frequencies)"

not correct, since the highest frequencies are in the *lower right* part. You can also point out that "higher freq=fine detail", that's a point that newbies find difficult to understand.

hope this helps :)
virus

[mikeX]

ok, here we go again :)

Spelling/Syntax:

B8: How do i use XviD on Linux?
B8: How do i use XviD on GNU/Linux?

it's not there. configuration
it's not there. Configuration

C4:
Please also realize that currently the actual limiter in the codec itself is not implemented yet,

C5:
could mention that AR is called DAR some times (Display Aspect Ratio?)

C8:
It is an avi file, and you can play it
- i see your point but it really depends on the encoding application (whether it is an avi file or not)
maybe: 'it is a playble video file...'

(Then you may choose to keep it, as the file created by the first pass will then be completely normal)

C9:
accross
across

- something i've been wondering: why isn't 0 the default for Overflow Control Strength?

I really like the 'Curve Compression Explanation' :D

C11:
settings..

maybe elaborate a bit on PSNR (at least the meaning of the term)

maybe also give the approximate min/max values for B-frame sensitivity (~ -35,+25 i think), i.e.:
"There are no real min/max values, but values under -35 and over 25 shouldn't make any more difference"

Weight zones with weight lower than 0.02 break filesize prediction.
http://forum.doom9.org/showthread.php?threadid=70379&perpage=20&pagenumber=4
unless sysKin made a typo on that post that's 0.2 (didn't get around testing it myself, only tested quant @ 31 which worked fine)

C12:
A)
how every pixels

Each frame a part of the picture where the block no longer is becomes the background colour while another part has to become the block colour. The change that would have to be encoded each frame would constitute a significant amount of bits.
:confused:

texturebits
texture bits ?

Setting 1 has a relatively small impact and it is recommended for all encodes.
maybe "A setting of 1 has a relatively small impact and is recommended for all encodes. "

maybe elaborate some more on 'Turbo', i.e. minor boost in speed, *minor* degradation in quality ?

C14:
Offcourse
find/replace all 'Offcource' && 'Off cource' with 'Of course' (damn and i was sure it was off :()

AMDs (at least Athlons) have SSE too, i think some latest models have SSE2 as well (not sure)

or DX50 to play with the DivX 5 codec.
maybe use 'decode'

C15:
did use them know and then

Although the option has been removed a long time ago, clips encoded this way somehow still work though.

with another QM work
with another QM works or may work

but are not
but is not

I discovered that a while ago that

D1:
even then, Hitting

D2:
sure all you're settings

make sure all you're settings are what they are supposed to
i don't think this means anything :confused: (maybe you meant "are what you wanted them to be?)

Switch to MPEG quantizer
"Switch to MPEG Quantization Matrix" (to avoid confusion)

Reduce the strenght

Also try more conservative B-frame settings.
maybe "Also try more conservative B-frame ratio/offset/sensitivity settings."

D7:
has enough power
better use "has enough cpu power", :D you never know what someone might think (what! do i need a new power supply :confused: )

D8:
information is encoded and decoded differently
i see what you mean but i think "information is decoded differently" is enough and more correct

E3:
the code and incompatabilites

Where not talking about security-patches
do you mean we're?

DShow filter are using to
DShow filter are you using to

Please post all your options you used
Please post all the options you used

kicked of that forum
kicked off that forum

Specify if your encode is muxed with audio (specify the audio) or not.
maybe add subtitles too(due to the vobsub flip-image thing)

E5:
least a garanteed level of
least a guaranteed level of

glory of XviD.And offcourse
glory of XviD. And of course

List of know bugs for different

-----------------
video - videos
oops, sory for the 'video' stuff, just couldn't think of an expression using 'videos' at the time (duh, music videos)

about B8:

Mencoder is not a standalone project. It's built into Mplayer --> www.mplayerhq.hu
http://www.mplayerhq.hu/DOCS/HTML/en/mencoder.html (mencoder's documentation)
adding mosu's 'dvd ripping and transcoding with Linux' guide may be a good idea too:
http://www.bunkus.org/dvdripping4linux/index.html
it's a bit outdated at the moment, will pm mosu to see if there are any plans for updates/linking to xvid faq etc

you may also wanna add this link:
http://zebra.fh-weingarten.de/~transcode/xvid4conf/
it's a program that resembles the xvid VFW gui and works with transcode's configuration files (and can also be used with dvd::rip --> http://www.exit1.org/dvdrip/)

For people wanting to mirror my FAQ or covert it to Pdf, Doc, Sgi etc., please follow the following guidelines:
what about translating?


well, the faq seems to be taking it's final form :)
My thanks to crusty and everyone who contributed for a REALLY GREAT XviD (and not only) FAQ ;)

[atreya2011]

Originally posted by mf
*VERY* simple explanation: the image is converted to detail signals (DCT), and then those signals are divided (coefficient cutting) to make it smaller. The more signal you cut off, the more detail you lose and more artifacts (blocks, ringing aka mosquitoes) you get. The quantizer is the detail removal factor (DivX3/Nandub thus calls quantizers DRF). The higher the quant, the more detail gets cut off, the lower quality.


Just a sugesstion.

Maybe you could add this in A4. MPEG-4 Basics section

Q. What's a quantizer?

[crusty]

Can't update my FAQ right now...The webserver is a bit boogered, and can't refresh the file..
I'll have to wait until it is fixed. In the meantime I'll include the error corrections in my version at home.
@virus:
"Of course", not "Offcourse"
I'll do a replace all.
A17:
same
B2:
Good point, I'll add it.
B5:
Ok, will look into it.
C3:
It's a quote from another document..hardly something I feel entitled to correct. 'If quoting, quote correctly'.
But what is probably meant is the frequencies for one axis only.
If you take into account both axes, then it would be in the lower right corner.
I'll probably add another link to my explanation of the process at:
http://forum.doom9.org/showthread.php?s=&threadid=54147
which explains a lot.

@MikeX

B8:
I think just Linux is good enough, after all everybody knows it by now. Let's try not to complicate things absolutely more than necessary OK? :) :D

C4:
Please also realize that currently the actual limiter in the codec itself is not implemented yet, as it's hard to do properly.
How about this:
"Please also realize that right now the actual limiter in the codec itself is not implemented yet, as it's hard to do properly."
(underlining not in the FAQ, just here to focus on the difference)

C5:
Good point, I'll add it.

C8:
Well usually the outcome is an avi file, but I'll change it to:
"It is a proper video file, and you can play it."

C9:
Another one of those document-wide replacements coming up... :)
something i've been wondering: why isn't 0 the default for Overflow Control Strength?
Uhmm..did I miss something? I was under the impression that it was... :confused:
Anyway I'll check it this evening.
I really like the 'Curve Compression Explanation'
I like it too. :D :cool: :D
Actually, the part I like the most is that I actually understand it at all...LOL!

In fact, during the writing of this FAQ I found that most options actually make good sense and are quite easy to understand. (Well at least to me)
The biggest problem for me was finding the right explanations in forum threads.
But, if you understand the basics of MPEG-4, everything makes sense much quicker and easier than if you don't have any clue about how it works.

Thinking about this makes me wonder if I should put a better and more fundamental explanation of MPEG-4 into the FAQ. I Probably should.

If you understand the basics behind motion search, psychovisuals (which I haven't adequatly explained at all yet) and the DCT-quantization process, things start making sense very quickly.

You can expect a major rewrite of some of the questions in section A in the future. Especially after Xvid goes 1.0, when it makes more sense to me to take out some of the items there.
(btw, looking at the (lack of) current bugs I expect the next build will be the latest RC, just to fix the really minor bus that are now present. So I guess I should start rewriting soon)

C11:
OK, I see it. Will update.

PSNR:
It's not really an XviD thingy, but expect it to be mentioned in a glossary some time in the future. Indeed, it's mentioned just too often to be overlooked.
@B-vop sensitivity:
I already put this in:
Use small values at first, it's quite sensitive.
But I'll ask sysKin for some more details on exactly how sensitive it is.

C12:
I hardly think 'texturebits' is a proper word anyway. :D
Will update.
Motion search is the process in which the codec is trying to figure out how every pixels of the original clip was moving.
Got the error.
Still, looking at it again makes me think it's not really a proper explanation. It's not the pixels that are moving, it's the content that gets portrayed by those pixels that is moving.
So it's not the motion of the pixels that gets captured, but the (virtual) motion of the objects that is captured. The codec couldn't care less about pixels anyway.
I'll have to rewrite this...
Setting 1 has a relatively small impact and it is recommended for all encodes. maybe "A setting of 1 has a relatively small impact and is recommended for all encodes. "

Well you can look at 'setting' as either a verb 'to set' or as a noun 'the setting'.
I chose the noun.
Turbo:
Indeed, forgot to mention it affects speed. Will update.
(btw:There is no *practical* degradation in quality AFAIK. But I don't know enough about it really to make a proper assumption about that)
AMDs:
I thought T-birds didn't have SSE, Athlons have SSE. Isn't SSE2 a PIV-only thing?

I would *love to* include a small table of cpu's and the instruction sets they have. If someone could post one here I'll include it asap.


C15 'know and then' in 'modulated QM' ...found it.
Anybody know any other non-compliancies?
'work' ...found it.
'but are not considered MPEG-4 compliant'..found it.
' discovered that a while ago that, while Qpel usually '...found it.
'Although the option has' ..etc. It's not exactly literature, but it's correct. ;)

D1:
found it.

D2:
'-make sure all you're settings are what they are supposed to, and that there is no error in any filter setting in avisynth and/or virtualdub.'

How about this:
"
-make sure all your XviD options have the correct settings
-make sure there are no errors in your filter settings and/or scripts in avisynth/virtualdub/whatever-tool-you-use."

"Switch to MPEG Quantization Matrix" (to avoid confusion)
Good one, will update.
Reduce the strenght
What's wrong with it? :confused:
maybe "Also try more conservative B-frame ratio/offset/sensitivity settings."
Hmmm.. I think that would be 'rubbing it in' :D

D7:
CPU power....good one. Don't want people blaming me for the wrong hardware upgrades. :D
The effect this has is that some colour information is encoded and decoded differently when playing an old file with a new decoder.
You're right, I should make it a bit more clear.
How about this:
"The 'Simple' process encoded colour information in another way than the 'Walken' process, and the latest decoders use the 'Walken' process to decode. Because of this discrepancy sometimes colour information might be decoded differently."
Much much better isn't it?
Also, the last line is incorrect, the 'bug' is known (not really a bug, more a result of a development decision) but there is no proper solution yet.
I'll make it this:
"The work on a solution to this issue is currently in progress. Since you can't detect the difference automatically, the decoder will have an option to switch decoding between 'Simple' and 'Walken'."

E3:
'incompatabilites'
replace all....
Where not talking about security-patches
Roger, will update.
-In Windows, what DShow filter are using to decode it
Not my error but Nic's. :D
Will update.
'Please post all the options you used'
Same...
'kicked off that forum'
Roger, will update.
'maybe add subtitles too(due to the vobsub flip-image thing)'
That's a good one! I think that deserves a separate question in section D.

E5:
'XviD would mean there's at least a garanteed level of continuous development.'
Roger, got it.
'glory of XviD. And of course '
..got it.
'List of know bugs for different builds and workarounds'
..got it.

B8:
Mencoder is not a standalone project. It's built into Mplayer --> www.mplayerhq.hu
http://www.mplayerhq.hu/DOCS/HTML/en/mencoder.html (mencoder's documentation)
adding mosu's 'dvd ripping and transcoding with Linux' guide may be a good idea too:
http://www.bunkus.org/dvdripping4linux/index.html
it's a bit outdated at the moment, will pm mosu to see if there are any plans for updates/linking to xvid faq etc
you may also wanna add this link:
http://zebra.fh-weingarten.de/~transcode/xvid4conf/
it's a program that resembles the xvid VFW gui and works with transcode's configuration files (and can also be used with dvd::rip --> http://www.exit1.org/dvdrip/)

Ok, I will give a link to both Mplayer and dvd::rip
The weingarten link is just a ftp directory...I dislike linking to those, as they are not really explanatory.

what about translating?
As long as it's not done with Babelfish, and you add a link to the official page it's OK. PM me with the location and I'll put it in a mirrors/translations appendix.

Thanks a lot for the corrections and keep it up!
(btw look at the bottom of the FAQ :D )

@atreya2011:
'Maybe you could add this in A4. MPEG-4 Basics section'
Sounds good, will look into it.

[crusty]

OK updated all mentioned errors (and some more).

I will have to look a bit longer into some of the other things tho..

EDIT:

Added:
D10. I installed XviD on XP but it doesn't show up in the codec list!

D11. My old XviD files don't play correctly with the latest XviD!

D12. Other issues...
(about Asian Windows, and about WMP9 issues)

[virus]

Originally posted by crusty

C3:
But what is probably meant is the frequencies for one axis only.
If you take into account both axes, then it would be in the lower right corner.

I'll probably add another link to my explanation of the process at:
http://forum.doom9.org/showthread.php?s=&threadid=54147
which explains a lot.


Your explanation in that thread looks good, but be aware that DCT stands for 'Discrete Cosine Transform' (not 'Cosinus', try a Google search with both expressions, you'll see how many pages turn up... I got 54000 vs 800 ;) ).
As for the frequencies... imho it's better not to quote text if it's unclear, I've seen some ppl confused on this topic here.

cheers :)
virus

[mikeX]

D10:
see the codecs files that are Windows is to use
see the codecs files that Windows is to use

shouldn't the question be 'i can't see xvid in the codec list on vdub(mod)'??
btw a good addition :D

D11:
There are many probably causes for this
There are many probable causes for this or
There are probably many causes for this

D12:
from a compatability mode.
you missed that one : P
ANYTHING BUT WMP9.
(:D you got that right)

I think just Linux is good enough, after all everybody knows it by now. Let's try not to complicate things absolutely more than necessary OK?
fine by me, just don't be surprised when those threatening 'Linux-->Kernel, GNU/Linux-->OS' emails start showing up :D)
seriously now, i see your point...

C4:
well actually i meant that you have both 'currently' and 'yet' (could drop the 'yet')

Uhmm..did I miss something? I was under the impression that it was...
nope, it's 5 (i think it has been since beta2 or something - not sure though)

yeap, you've got that right about the basics of MPEG4, i believe that's what makes your FAQ different from other FAQs, and quite more useful and comprehensive overall
I don't know if that's the same from the point of view of a total noob though, I mean i'm no video expert myself, but i learned a lot of stuff in the last few months about video encoding etc, most of which are present in your FAQ.
From such a point of view i find it very comprehensive and all-inclusive, but some feedback from someone clueless (but eager to learn) would be great!

(btw, looking at the (lack of) current bugs I expect the next build will be the latest RC, just to fix the really minor bus that are now present. So I guess I should start rewriting soon)
looks like it, but no need to rush things, quality over quantity ;)

C11:
About the b-frame sensitivity, it's sysKin i'm quoting on those values, but they are not 'absolute' values so it's kinda tricky whether to put them or not.
Most people would probably feel 'relieved' to see a max/min value though (what can i say, people want to feel restrained after all :rolleyes: )

C12:
Yeap i see your point about the pixel thingy, could also cause confusion with q-pel

Well you can look at 'setting' as either a verb 'to set' or as a noun 'the setting'.
oops, that's why it didn't make sense, didn't even cross my mind :D

"make sure.."
sounds good :)

strenght
should be: strength

D8:
great!

The weingarten link is just a ftp directory...I dislike linking to those, as they are not really explanatory.
yeah, see your point, it's mentioned in the dvd::rip INSTALL so it's ok, maybe it could be integrated to xvid in the future...


hehe, big all Mr. Contributor ;) thanx

translation probably on the works but at a slow pace (i posted on greek.doom9 about it)

[atreya2011]

@MikeX
Man, You must be the fastest gun in the west or something, I mean, I figure almost all those mistakes you do and then I say to myself, ok lets go post it and BINGO!, I see your post and then I say to myself, oh man back to the drawing board :D

@Crusty
By the way, how often should I add the link to PDF file in this thread as the FAQ is getting updated on a daily basis.

[mikeX]

@ atreya2011
:D the art of nitpicking...

@ crusty
might wanna check this out:
http://www.tommesani.com/InstructionSetCPU.html

and this as well (mentions sse2 support for AMD 64bit):
http://www.tech-report.com/reviews/2003q3/athlon64/index.x?pg=1

[crusty]

@virus:
Where did you find the word 'Cosinus' I can't find it.

@MikeX:

D10, D11, D12:
updated...

C4:
Well not anymore now..
nope, it's 5 (i think it has been since beta2 or something - not sure though)
I'll look into it.
looks like it, but no need to rush things, quality over quantity
My feeling exactly. I just think it's looking very good right now.

C11:
I'll look into it.

strength
'Replaced all'...

Hmm...I'll check these links.
http://www.tommesani.com/InstructionSetCPU.html

http://www.tech-report.com/reviews/...64/index.x?pg=1

Now how did I do tables again....hmmm (www.htmlgoodies.com here I come) :D

More cpu links:

http://www.qvctc.commnet.edu/classes/csc277/i_sets.html

Thanks for the corrections again.

BTW, I'll also update the Qpel question soon.

[Soulhunter]

Really not want to add it (http://forum.doom9.org/showthread.php?s=&postid=443398#post443398) ???

Or should I write a summary about this stuff... :p

Bye

[mikeX]

Really not want to it ???
oh, i wanted to point that out again (but totally forgot about it)
i find it a useful addition, and i remember it gave me some nasty headaches before i came across those posts....