Doom9's own words..... and even without his articles I would choose Nandub :o) [Archive]

NeVeRLiFt

29th December 2001, 19:14

Conclusion

At 900kbit/s the choice was clear: SBC wins, even with unoptimized settings. At 1800 it's already a close call and at 6mbit/s I wouldn't want to appoint a winner.

Ya when you know how to optimize it Nandub(SBC encoding) get even better ;-)

http://www.doom9.org/codec_1800.htm

If you have a great-looking SBC rip of a certain title redo it in DivX4 and compare the results. If you find problematic areas let us know via the DivX encoding forum.

Doom9 I have lots of great SBC encoded movies. And when I try DivX4.xx the quality is ok. I will stick to the old DivX since it does not give me problems, have bugs or crashes. I have no problems with the old DivX.
I wish the guys making DivX4.xx the best of luck!
And what your doing is great :D

Have A Great HAPPY NEWYEAR!!!

PS:
Even without Doom9's article I would choose Nandub(SBC encoding)
I know how to use it and I have seen first hand the quality and size it can do. I dont need a easy way to encode! I want the best and at this time SBC Nandub kicks ass. Maybe someday DivX4.xx VBR will get there in terms of quality. But then what about the bugs and crashes and weird problems it has. I dont have them problems with the old DivX.

ProfDrMorph

29th December 2001, 23:28

With DivX3.11 freeze problems I actually had more problems with the old DivX than I ever had with the new one. IMO SBC is not better it's as good as DivX4 2-pass VBR.

NeVeRLiFt

29th December 2001, 23:36

Originally posted by ProfDrMorph
With DivX3.11 freeze problems I actually had more problems with the old DivX than I ever had with the new one. IMO SBC is not better it's as good as DivX4 2-pass VBR.

Knock on wood! But I have never had freeze problems. Maybe your doing something wrong? You should not gripe about DivX's you have downloaded(not saying you are....). Its just alot of n00bies download DivX movies and then complain, they dont know what there doing and a freeze frame can be fixed in just a few seconds anyway.

-h

30th December 2001, 01:05

If you want the reason for ms-mpeg looking better than divx4 (and it does), it comes down to the motion estimation performance. That is, MS-MPEG43 does a better job of finding similarity between images, and can thus fit the same quality video (i.e. same quantizer) into a smaller space.

Here are the fixed-quantizer results from a test I did (K7 1.2 Ghz) - the first 3175 frames of PAL Crouching Tiger Hidden Dragon VTS_01_3.VOB with Crop(0,70,4,71) and BicubicResize(720,320):

time taken (decompression time subtracted):
quant divx4 mp42 mp43 xvid h xvid m
02 91 103 102 105 148
03 84 92 91 99 142
04 80 84 83 95 137
05 78 83 82 94 133
06 76 81 79 92 132
07 74 79 78 91 130
08 73 77 76 90 130
09 72 76 75 90 129
10 71 75 74 89 128
11 69 75 74 88 127

file size:
quant divx4 mp42 mp43 xvid h xvid m
02 65802240 61507584 60897280 58832896 60157952
03 42319872 40513536 39952384 38270976 35454976
04 30179328 26359808 25702400 24905728 25630720
05 23492608 20838400 20252672 19380224 18827264
06 17256448 15826944 15214592 14446592 15040512
07 14571520 13494272 12908544 12183552 12148736
08 12460032 11065344 10475520 10856448 11503616
09 10354688 9854976 9275392 9670656 10203136
10 9226240 8470528 7905280 8337408 9105408
11 8343552 7774208 7208960 7628800 8157184

Just to clarify, divx4 is DivX4 4.12, mp42 is microsoft's "other" MPEG4 codec (which disables the prefilter?), mp43 is the usual DivX 3.11 codec, xvid h is XviD (built from 01-12-27 CVS) with the H.263 quantizer, and xvid m is the same XviD CVS build but using the MPEG quantizer (sharper images).

So looking at that, with MP43 (or MP42) you get the same quality as DivX4 but with a smaller file size - thus a 1-CD rip can contain more image data, thus regardless of the encoding method used, MP43 is *empirically* a better format at the moment. That said, XviD is currently better than all of them until you get to quant/DRF=8.

As a side note, MP43 is actually cheating, as it prefilters the image to make it more compressible (given motion thresholds), though I admit it is hard to notice. If it was possible to disable this filter, you would see file sizes rise to levels probably the same as DivX4. It appears a much stronger filter is used when quant >= 8 also. It is my assumption that this filter is disabled in MP42, though I have never heard any .dll hackers mention the specifics.

-h

NeVeRLiFt

30th December 2001, 01:25

I did not know all that.... but thats why I come here, to learn!
All I do is test and use what works best. I did a rip of Crouching Tiger Hidden Dragon and it was very small and still had impressive quality! Its a dark movie. The file size was 618mb and CBR mp3 ;)
I took some of my DivX movies to work and also some VCD's I had made using TMPGEnc and showed my boss. He was very impressed and now where gonna start a small business doing movies and digital pictures for various occasions(Weddings, Bdays etc) He even bought a DVD recorder and DVDit LOL and very syched! Thanks goes out to Doom9 and Nicky and VCDhelp.com :D I learned it all from you guys!

btw
I build and repair computers for a living.

You really wanna test the DivX codecs do video capture with them!
Then you will see..... the VKI mpeg4 V3 codec kicks ass for Video Capture, I'm talking filesize of 5-12mb per 1min and quality of a SVCD!

yangus

30th December 2001, 09:43

-h

30th December 2001, 11:29

> so is this why divx 3.11 have a hard time with scenes containing lots of actions?

MPEG-based codecs work by storing the difference between subsequent frames - the bigger that difference is, the more data has to be stored. Sometimes, high-action scenes won't incur a large bitrate, as the action on the screen may lack intricate details (detail takes more bits to encode than plain areas).

On average, however, the more action that is occuring in a frame, the higher the cost in bits will be. There are ways to bring this value down (better INTRA/INTER MB choice, or even brute forcing it), but it brings a large cost in compression time. XviD may have this option in the future, as it significantly increases the compression ratio of high-action soruces like anime.

-h

ProfDrMorph

30th December 2001, 19:34

Knock on wood! But I have never had freeze problems. Maybe your doing something wrong? You should not gripe about DivX's you have downloaded(not saying you are....). Its just alot of n00bies download DivX movies and then complain, they dont know what there doing and a freeze frame can be fixed in just a few seconds anyway.
I did the rips myself. I'm complaining about the fact that these frames may occur even though this problem can be easily fixed. With this minor problem I had exactly one minor problem more with DivX3.11 than with DivX4.x which worked without problems for me so far. I still prefer DivX4 since it's a lot easier to use. I'm only doing 2-pass rips and then I don't need to set up Nandub/SBC. DivX4 gives very good results, too. Only if there would be a huge difference ( not 5mb as you can see in h's comparison but 15mb or more ) then I would step back to the ILLEGAL codec again. But at the moment DivX4 even though not being the best codec is good enough and it doesn't make you think about ethics.

NeVeRLiFt

30th December 2001, 23:49

Well its only a matter of time now before DivX4.xx becomes the Codec to use. They keep working on it making it better..... but there are some bugs that need to be worked out, that being the case even if it is better I want use it till the bugs are fixed.

-h

31st December 2001, 00:58

> Only if there would be a huge difference ( not 5mb as you can see in h's comparison but 15mb or more ) then I would step back to the ILLEGAL codec again.

Actually, given that I only did a 3175-frame segment of the film, for the entire flick you'd see sizes like:

file size (MB):
quant divx4 mp42 mp43 xvid h xvid m
02 2510 2346 2323 2244 2295
03 1614 1545 1524 1460 1353
04 1151 1006 980 950 978
05 896 795 773 739 718
But that's just multiplied out - I'm not encoding the whole movie that many times ;) When you look at full-length films, quite a large difference appears.

Of course if you were encoding a movie with XviD, you'd use lumi masking, which would drop the file size by a further 20%.

-h

ProfDrMorph

31st December 2001, 03:15

nice comparison h. But everybody be honest: if you know how to do DVD Rips right it doesn't matter if you're using DivX3.11 or DivX4 for two CD rips, does it? XviD surely will beat both DivX3.11 and DivX4 but it still needs to smooth out the bad code of OpenDivX and the DivX4 parts that once were open source.

NeVeRLiFt

31st December 2001, 03:21

I have never had any problems making good looking 2cd rips.
What I was refering to and defending was how good a 1cd rip Nandub can make. Thats all and this stupid ass everwicked just not understand!

-h

31st December 2001, 05:01

XviD only has a few hundred lines of OpenDivX code left - the rest has been rewritten. There's only one known image corruption bug left (pure vertical motion), then there's just optimizing and featuritis to go.

I just wanted to point out how the codecs perform against each other, in order to achieve identical output quality. Of course that's subjective though - MP43 smears the image much more than I could stand at higher quantizers.

I'm not particularly interested in backing up DVDs, I just like the technology involved :)

-h

ProfDrMorph

1st January 2002, 22:51

XviD only has a few hundred lines of OpenDivX code left - the rest has been rewritten. There's only one known image corruption bug left (pure vertical motion), then there's just optimizing and featuritis to go.
I know that. I follow the discussion on www.videocoding.de, too. Even though I nearly never post anything. I were refering to the image corruption bugs. I just read the latest posts five minutes ago and it seems that there may be more than just one bug that causes problems. Isibaar thinks ( and hopes ) only gruel's new ME is the cause.

-h

1st January 2002, 23:18

Yeah gruel's new code is very interesting, but still buggy. Pity, because PMVfast was giving equal file sizes and running 8% faster. The crud that's being encoded now is probably raising file sizes as well.

Using the latest build with the old ME still works fine. I wonder how many of XviD's features will be reimplemented in DivX 4.5..

-h

ProfDrMorph

2nd January 2002, 03:12

Using the latest build with the old ME still works fine. I wonder how many of XviD's features will be reimplemented in DivX 4.5..
I'm sure they try everything on which they find information. But I don't think that DivX will beat XviD in the future since DivXNetworks do it for money. They aren't interested in a good codec but in making it look good. XviD is different. The developers ( everybody can help developing XviD! ) do it because they want efficient techniques. This different aim makes the XviD developers work harder than the DivX ones and so XviD will be better than DivX.
Ok I can't say it it's really like that but at least I think this comes close.

BetaBoy

2nd January 2002, 06:26

Nice thread but a few things....

@ NeVeRLiFt - SBC is a method, a software tool not capable of encoding 4.xx like 3.11... lets see a compatible 4.xx version of Nandub or the like and get back to me what version Doom9 would like..

@ -h , Can you tell me if XviD is going to be a proprietary codec OR MPEG-4 based and if it is MPEG-4 based who will pay for the IP?

@ -h - Agreed, motion estimation performance is an issue that is being addressed and will not be an issue soon ;-)

@ ProfDrMorph - The QUALITY and compatibility is at the TOP of what we strive for. To think we have stopped the dev process with the 4.xx line is to be foolish. We have just begun. 2002 will be an even bigger year with all the things planned for the future v5.0 and v6.0 versions of the DivX codec.

BlackSun

2nd January 2002, 07:37

I want BetaBoy 6.0, it the best :p
When DivX 4.5 will be available ?

-h

2nd January 2002, 07:37

> Can you tell me if XviD is going to be a proprietary codec OR MPEG-4 based and if it is MPEG-4 based who will pay for the IP?

No one's supposed to use XviD, it's just there for educational purposes, like that LAME program that no one's ever made an mp3 with..

Seriously though, I don't think it's too big of an issue, especially with tools as popular as DVD2AVI, mpeg2avi, Flask, XMpeg and azid all being illegal (with regard to licensing) as well.

-h

ChristianHJW

2nd January 2002, 10:35

@h :

Is it correct that XViD is still using DivX4 DirectShowFilter for playback ?

If so guys, lets be honest and we all should admit that there was quite some work being invested from DARC in getting this DSF to work, and its doing a wonderful job now, even with DivX3 movies and on low power CPU's ......
Just remember all the hardware compatibility issues they had to go through ..... will be an interesting time for XviD, once you'll develop your own DSF ...

-h

2nd January 2002, 11:57

> Is it correct that XViD is still using DivX4 DirectShowFilter for playback ?

XviD isn't using any dshow filter at the moment, it was abandoned when the new vfw frontend appeared. XviD's decoder (again, only a few 100 non-gpl lines remaining) is faster than DivX4's at the moment, especially on low-end pc's, which the developers all seem to have! It still calls divx4's dshow filter if you play with wmp though, but I disable it.

It's on the todo list.. Isibaar's working on pre/post processing code, and since I found a wonderful chap willing to play with the MacOS QT port, I might have a stab at the dshow component.

-h

BetaBoy

2nd January 2002, 15:32

Originally posted by -h
> Can you tell me if XviD is going to be a proprietary codec OR MPEG-4 based and if it is MPEG-4 based who will pay for the IP?

No one's supposed to use XviD, it's just there for educational purposes, like that LAME program that no one's ever made an mp3 with..

Seriously though, I don't think it's too big of an issue, especially with tools as popular as DVD2AVI, mpeg2avi, Flask, XMpeg and azid all being illegal (with regard to licensing) as well.

-h
@-h - So I guess you mena it will be MPEG-4 based. If so, once an XviD 1.0 release is available it will be able to properly parse an MP4 stream?

btw... I did not think it was a big deal to distribute 3.11 on MyDivX.com... that was till MS called my house and sent me some mail with some crazy high 5 figure $$ amount in possible fines (never mind legal costs) asking me nicely to remove it ;-(

Thats what popularity drives... I hope for XviD sake that development gets alot farther along before something like that ever happens.

gldblade

2nd January 2002, 23:02

>No one's supposed to use XviD, it's just there for educational purposes, like that LAME program that no one's ever made an mp3 with..

LOL :) Same thing with DivX 3.11. Did LAME ever get into any trouble?

-h

2nd January 2002, 23:17

> So I guess you mena it will be MPEG-4 based. If so, once an XviD 1.0 release is available it will be able to properly parse an MP4 stream?

I don't believe there are any plans to support MP4 system streams, just the video profile(s). MPEG4IP is looking at implementing XviD as its video codec, and BlackSun tells me PowerDivX is going to natively support MP4 stream playback in the future. No I don't know what he's going to do about the licensing either ;)

Any legal issues will probably find their way back to Isibaar - though considering LAME, FAAC and MPEG4IP have been able to go about their business free from (enforcable) legal threats, I'd say the same should hold for XviD. Time will tell.

-h

ProfDrMorph

3rd January 2002, 03:28

The QUALITY and compatibility is at the TOP of what we strive for. To think we have stopped the dev process with the 4.xx line is to be foolish. We have just begun. 2002 will be an even bigger year with all the things planned for the future v5.0 and v6.0 versions of the DivX codec.
I don't think that DivXNetworks has stopped development. What I wanted to say is that I think the open source project will beat the commercial closed source project since people working on open source projects don't do it for money and so they're more motivated than others. Ok surely there are people who earn money for something and work highly motivated on it but often commercial software can be so crappy compared to free software. E.g. take Game Voice and BattleCom. Especially in the beginning there were many complains in the Game Voice forums with requests by people asking for features that GV lacks/lacked but BC has.

btw: good to read that you defend DivX that strong. For me this indicates that DivXNetworks doesn't count to those people doing a crappy job for their commercial product.

BlackSun

3rd January 2002, 07:30

Originally posted by -h
> So I guess you mena it will be MPEG-4 based. If so, once an XviD 1.0 release is available it will be able to properly parse an MP4 stream?

MPEG4IP is looking at implementing XviD as its video codec, and BlackSun tells me PowerDivX is going to natively support MP4 stream playback in the future. No I don't know what he's going to do about the licensing either ;)

-h

Yeah I am interested in supporting MP4 stream playback because it one of the usage of MPEG-4 codec, I mean streaming. Now my problem is: should I wait for DivXNetworks to release a DirectShow filter for MP4 playback, or maybe they will simply go into trouble, and just allow The Playa and DivX Player to play MP4... Actually I dunno what to do, licensing is still the main problem for me (even if I need more coders for MP4 playback). Even if DivXNetworks don't answers my mails anymore, my player is still called PowerDivX and still interested in streaming/VOD :)

NeVeRLiFt

4th January 2002, 16:35

Originally posted by BetaBoy
Nice thread but a few things....

@ NeVeRLiFt - SBC is a method, a software tool not capable of encoding 4.xx like 3.11... lets see a compatible 4.xx version of Nandub or the like and get back to me what version Doom9 would like..

Man I still would not use DivX4.xx it has to many bugs. Sorry but I dont have the same problems with 3.11a or VKI mpeg4 V3 codecs.
This is not against you dude, I'm just tired of lamers running there mouth and not admiting that Nandub does a good job and its been shown over and over again. Some lamers like everwicked still dont get it :rolleyes:

cofferscuffs

4th January 2002, 16:53

Deja Vu? I could of sworn you posted this on Rage3D :D BTW Blacksun, I wouldn't wait for DivX Network's streamable codec - I still prefer RealVideo 8 (please don't flame me), as it's proven to be the best at low bitrates (even though the speeds of the internet are getting faster and faster, and with IPv6 coming up....). XviD is potentially going to be one of the most 'catalistic' codecs for this year - why? It will force DivX Network to work harder as XviD didn't go commercial (I've still got grudges for that). I however, still believe that VOD systems are not viable, and do not offer any advantages of going to the cinema/PPV/renting videos. I'd rather spend my £3 on renting a DVD, then getting a low-quality movie over the internet which delivery depends on speed. Bring on XviD!! Just my 2$.

BetaBoy

4th January 2002, 18:29

@cofferscuffs we don't need XviD for that. WFor myself wrking 20 hours a day for the past year has been competition enough with what's out there already.

I wish XviD good luck... As usual, I am think as a way of a project manager and what steps are needed for it to become a popular codec without the hastles in the future.

With DivX 3.11 it was only one (1) comnapny... MS. Now with MPEG-4 you have nineteen (19) very powerful companies and as many as a dozen others as well you might potentially have to deal with for license policeing. Though MPEGLA has reduced this to a simple one-stop shop license. It remains to be seen that if interest continues towards the movement of MPEG-4 and XviD becomes popular. It might become a bigger issue.... thats all.

-h

5th January 2002, 01:08

I don't really think DivXNetworks needs XviD for competition - with licensing worked out, many more MP4 creation utilities will appear - I know Ivan of Psytel fame whipped up an MPEG4 video encoder in astonishing time (with quite nice results), and more are on the way. QuickTime 6 for example.

It'll be very interesting to see the eventual crop of .MP4 creation utilities compared actually.

-h

AM Tiainen

5th January 2002, 18:20

Originally posted by NeVeRLiFt
Conclusion

At 900kbit/s the choice was clear: SBC wins, even with unoptimized settings. At 1800 it's already a close call and at 6mbit/s I wouldn't want to appoint a winner.

Ya when you know how to optimize it Nandub(SBC encoding) get even better ;-)

http://www.doom9.org/codec_1800.htm

If you have a great-looking SBC rip of a certain title redo it in DivX4 and compare the results. If you find problematic areas let us know via the DivX encoding forum.

Doom9 I have lots of great SBC encoded movies. And when I try DivX4.xx the quality is ok. I will stick to the old DivX since it does not give me problems, have bugs or crashes. I have no problems with the old DivX.
I wish the guys making DivX4.xx the best of luck!
And what your doing is great :D

Have A Great HAPPY NEWYEAR!!!

PS:
Even without Doom9's article I would choose Nandub(SBC encoding)
I know how to use it and I have seen first hand the quality and size it can do. I dont need a easy way to encode! I want the best and at this time SBC Nandub kicks ass. Maybe someday DivX4.xx VBR will get there in terms of quality. But then what about the bugs and crashes and weird problems it has. I dont have them problems with the old DivX.

What????
My opinion is that old DivX 3.11 is worse than DivX4. DivX 4 make better quality, it is clearly, never block very much and it never make freeze frame. I have also try DivX 3 and DivX 4 to make DVD rip and usually DivX 4 make better quality. And now when is possible to use DivX 4 Log file analyzer (http://divx4log.narod.ru/)nobody can say that divx 3 can be make better quality even you optimize nandub options. You can also look updated DivX quality compare http://server1.doom9.org/codec-comparison-update.htm. And they test is maked with divx 4 wins even with unoptimized settings.

NeVeRLiFt

6th January 2002, 04:17

Originally posted by AM Tiainen

What????
My opinion is that old DivX 3.11 is worse than DivX4. DivX 4 make better quality, it is clearly, never block very much and it never make freeze frame. I have also try DivX 3 and DivX 4 to make DVD rip and usually DivX 4 make better quality. And now when is possible to use DivX 4 Log file analyzer (http://divx4log.narod.ru/)nobody can say that divx 3 can be make better quality even you optimize nandub options. You can also look updated DivX quality compare http://server1.doom9.org/codec-comparison-update.htm. And they test is maked with divx 4 wins even with unoptimized settings.

The page cannot be found
The page you are looking for might have been removed, had its name changed, or is temporarily unavailable.

--------------------------------------------------------------------------------

Please try the following:

If you typed the page address in the Address bar, make sure that it is spelled correctly.

Open the server1.doom9.org home page, and then look for links to the information you want.
Click the Back button to try another link.
Click Search to look for information on the Internet

Darksoul71

6th January 2002, 12:49

Hi !

This might interessting for this discussion:
I´m just reencoding "Verrückt nach Mary" (Sorry, I dunno know the english title of this movie). Source is a 4 MB/s MJPEG capture from VHS.
Datarate:
753 kBit/s (DivX 3.11)
771 kBit/s(DivX 4).
Target res: 448x320

OK, the results:
DivX3.11 SBC pretty good movie (you have to keep in mind that is from analogue source. So it´s a little bit "unsharp" as usual)
DivX 4.12 2Pass mode with default quantizer and no log file tool used:
Crappy video quality. Somewhat of the before discribed "smoothed macroblocks". I´m attaching a screenshot from this.

My two cents on DivX 4.12 2Pass vs. DivX 3.11 SBC :
On 2 CDs it doesn´t matter what codec you use. For 1 CD movies (be it from analogue or digital source) Nandub/GKnot rules. This might change in future if we get better readable log tool and I could bet that GKnot will support DivX4(5/6 ?) Logfile analyzing and recalculation for future.

-D$

pcsl999

6th January 2002, 14:16

Sorry, idon't think so.

Doom'9 samples were still picture; a very small part of movie.
I would talk about the whole movie.

3.11 have wrong base rule in software design for movie encoding.
Their design were wrong in the 2 codes ( Low motion & High Motion).

Low Motion quality:
Very good in still picture & low motion
very bad in high motion

High Motion quality:
Very good in high motion
bad in still picture & low motion

A action film have 60-70% in high, 40-30% in low motion
A love story have 30% in high, 70% in low motion
Nobody can take a right choice between low or high.
SBC choice the more & forget the less.

Prizm

9th January 2002, 11:09

After that quality comparison page on Doom9, you 3.11 users
are rather blind and stubborn to continue saying 3.11 is
the best above all.

The DivX 4 pics were sharper and contained less blockiness.
There may not have been much of a difference on some screenshots,
but if there were noticable differences, it was DivX4 that came
out better.

"But they're only still shots!"

err...that's what movies are made of. The less artifacts per
frame, the better the overall quality. (well, *that* was a revelation)

I've tried both codecs, and DivX4 comes out best...many times
allowing extra space to up the sound quality.

I personally have never had any problems with DivX4 (Bugs? Never
encountered any), but if you have problems - just use 3.11 and
optimize it as best you can.

If you don't *personally* have problems with DivX4 - then don't
stubbornly continue using 3.11 just to prove a point.

Prizm

Awake

9th January 2002, 11:43

@Prizm: Could not agree more.

I must have killed 50 hours or just trying to get a movie that didn't have blocks or lack of detail.

Quality is my only concern, so zero macroblocks, errors, skipping, etc.

SWORDFISH, for example, is 2.4 gigs when done @ 720x304 with a frame of 720x416 @ 3000kbs in 2 pass Divx4.12 using only Vdub (after DVD2AVI & VFAPI of course) with min+max quantizers @ 2, and is absolutely perfect.

The frame itself takes a negligable amount bits away from the movie IF ANY. It blends seemlessly with the rest of the black screen in fullscreen viewing mode. No pixels or evidence of encoding at all, like the slight noise you see in a black area within a film even if watching a DVD.

This is after trying all the various methods and fiddling with all the setting countless times at bitrates from 1500 to 6000. Even @ 6000 the others can't compare. Div3 lacks detail overall, especially with areas of uniform color, like stone pillars for example. But in Divx4.12 I can see the texture of the stone, hairline cracks, etc. But without the artifacts or blocks I notice in Div3 even @ 6000 kilobits. This is with multiple sources too btw.

When using Ndub SBC and min+max DRF 2, its fairly close, but still not as good and there are error blocks here and there that can only be removed with anti-shit, defeating the purpose of min+max DRF 2 for the maximum quality.

They are using AC3 to btw.

ChristianHJW

9th January 2002, 13:26

I read somewhere that M$ implemented a fixed video filter in MPEG4V3, improving compressability, but removing picture details .... dont know what's behind that and how they could implement this in the codec itself though ...

BlackSun

9th January 2002, 13:47

It has been discuted on www.xvid.org

ProfDrMorph

9th January 2002, 15:55

SWORDFISH, for example, is 2.4 gigs when done @ 720x304 with a frame of 720x416 @ 3000kbs in 2 pass Divx4.12 using only Vdub (after DVD2AVI & VFAPI of course) with min+max quantizers @ 2, and is absolutely perfect.
Two frame sizes? I think you made a mistake here ;). Btw: when you're using fixed quantizers ( both at the same value ) 2-pass makes no sense at all. You better use the quality mode and set it to 100%.

like the slight noise you see in a black area within a film even if watching a DVD.
This only indicates that the movie is from an analogue source and the ones who encoded it haven't used a noise filter. This is NO encoding error. But it causes errors when there's too much of it in the picture ;).

Foo

9th January 2002, 22:19

This is all well and good, but it is basically a bunch of crap :). Yes SBC is great. Yes DivX4 two pass is great. Each one has its place, and each one gives better results for certain titles. To say that one is absolutely better than another in every case is asinine. It's just a tool, use which one works best for a particular movie.

foo

Awake

9th January 2002, 23:50

Two frame sizes? I think you made a mistake here . Btw: when you're using fixed quantizers ( both at the same value ) 2-pass makes no sense at all. You better use the quality mode and set it to 100%.

Sorry, I should have been more clear. In order to ensure the correct aspect ratio, the movie is cropped @ 720x304. Then I use the 'Expand frame and Letterbox' option @ 720x416 so I can use clips from various sources and append them together. Perhaps this is not needed, but it sure seemed to me that if I just cropped it to 720x416 that things didn't look right. As for using Quality mode, the macroblocks are terrible, even at a bitrate of 6000 they are all over the place. It doesn't even come close.

This only indicates that the movie is from an analogue source and the ones who encoded it haven't used a noise filter. This is NO encoding error. But it causes errors when there's too much of it in the picture .

By the 'frame itself', I meant the added letterbox frame, not the individual 'frame' of the movie. I never said it was an error on behalf of the codec. Analog or digital, there is always a slight difference between black areas of a movie and the letterbox frame, even if its so small you have to look for it on your monitor, or even magnify it in order to see it. What I meant was, it appears to be a nullified area.

To say that one is absolutely better than another in every case is asinine. It's just a tool, use which one works best for a particular movie.

As I said, this is with multiple sources, such as FILM, PAL, and Interlaced NTSC. I even encoded entire movies and watched them completely after using various techniques. I didn't limit it to just clips from movies and trailers.

Let me tell ya, I am sick of watching them repeatedly to scrutinize them. :(

ProfDrMorph

10th January 2002, 14:22

As for using Quality mode, the macroblocks are terrible, even at a bitrate of 6000 they are all over the place. It doesn't even come close.
Then you've misunderstood the aims of qualitybased encoding: If you're able to set a bitrate ( the box is not grayed out ) then you aren't using quality-based encoding.

And what I also said was that if you're using fixed quantizers ( min = max ) then 1-pass encoding gives you the exactly same result as 2-pass encoding.