Why does 4.11 sux so bad compared to 3.11?

[gorg777]

I have been knockin myself out to figure out why anybody use the 4.xx series at all...The 3.11 quality blows it out of the water...What surpises me is that even with 2-pass...4.11 looks all blocky and chunky. with bleeding from block to block...I played with this for months and reformatted many times thinking it must be me....anybody else have the same experience?

[Gin_Quaker]

ok u must be doing this badly. I know i am a novice to divx but i have encoded many movies, the latest being my ghost in the shell dvd and it turned out exactly the same as the dvd i am not shitting you! Its also on 1 cd! now the reason why anyone (or me)uses divx 4 is because it is legal, i mean i have goten the same quality with it and since it has no illigal code i use it more than 3.x!

Just so people know i did it with the sbc rip pack, but instead of nandub i use virtual dub with divx 4.11 and added vbr mp3 with nandub.

[murattttt]

I have tried and upgraded to the newest divx and used it to the extrme. Even open divx codecs and that means encoding at least three movies with each of them. Among other divx codecs 4.11 is I believe the best one both in qulty and compressability. I am afraid you should consider some extra things compared to 3.11 as follows:
-Before installing a newer codec remove all of the previous divx.
-4.11 requires at least 100 Ghz or more computer power.
-4.11 is more dependant on a clear system set up.
I would write more however I think you will get many other messages and consider then.

[gorg777]

But still I get these blocky shapes almost 1/2 an inch wide that are all over the screen, with little bleeds..I have never been able to get rid of them...I have been using nandub but played with virutual dub last night...I must say i did use 3.11 for a rip of fight club last night and on 2 cd's it really incredible..I HAVE noticed that resetting registry values can be very difficult using divx 4.11
I will reformat my other drive tonight and try it using the 2pass method....Oh by the way, I am almost certain that Nandub automatically reverts you back to 3.11....it can't use 4.11 at least thats what it says on ....hmmmm not sure but I know i read that last night...think it on the nandubs own docs...GORG
ohh forgot
my sys:
AMDxp running at 1760mhz
368mb DDR ram running cas2 160 fsb
2 20 gig maxtors(no raid)
win2k service pack 2 professional

[ohliuv]

the only problem I have with DivX4 is the size prediction.
Quality for me is the same as DivX3.

[Apfelstruhdl]

I am almost certain that Nandub automatically reverts you back to 3.11....it can't use 4.11 at least thats what it says on ....hmmmm not sure but I know i read that last night...think it on the nandubs own docs...GORG

If u want to use 4.11 u should use vdub as nandub is only for sbc, i think. So try vdub and results should be better. 4.11 has also faster encoding speed on P4 or athlonxp systems and the version which is also Thunderbird optimized will follow soon. divx4 has much better decode options, too. U can use it for divx3 decode too as far as i know. But i heard that a good sbc is about the divx4 qualaty, but i hope divx4 will get better and better.

[NeVeRLiFt]

Originally posted by Gin_Quaker
ok u must be doing this badly. I know i am a novice to divx but i have encoded many movies, the latest being my ghost in the shell dvd and it turned out exactly the same as the dvd i am not shitting you! Its also on 1 cd! now the reason why anyone (or me)uses divx 4 is because it is legal, i mean i have goten the same quality with it and since it has no illigal code i use it more than 3.x!

Just so people know i did it with the sbc rip pack, but instead of nandub i use virtual dub with divx 4.11 and added vbr mp3 with nandub.

DivX4.11 does not work with Nandub. And if you want great quality 1cd rips nothing beats Nandub! DivX4.11 might come close or you might say tie Nandub.... but to my eyes Nandub is king.
Hell anyone can make a good 2cd rip but making a good 1cd rip takes something special! And Nandub has it ;)

[pale]

I have to disagree with NeVeRLiFt. In my book, DivX 4.11 beats 3.11 in most cases, and is on par with the rest. Try http://divx4log.narod.ru/ for quality and size predictability. And 4.11 seems to handle low bits/(pixel*frame) values much better than 3.11 did, so there's no need to resize (as I consider that the best way to resize is not to). DivX 4.11 movies tend to be more blurred, so if you resize, always use sharp bicubic.

[MoRRiX]

I have seen movies Encoded with 3.11 and have worst quality,also 3.11 is slower than 4.11 and my size predictions with 4.11 are 1Mb-Accurate.No problems with this codec.

[NeVeRLiFt]

I dont see anyone using DivX4.11 making great 1CD rips. They all use Nandub. And I have yet to see a DivX4.11 beat Nandub for making a 1CD rip. And for 2CD rips Nandub still has the best quality IMHO. I guess you guys use DivX4.11 cause its legal? I cant see using it for any other reason!

[unplugged]

In my experience:

DivX 3.11 (with SBC):
+ respect very well "pixel" detail in areas where colors have similar tonality (and also in poor detail/contrast images, ...not crisp movies...)
+ conseguences: DivX 3.11 encodes with fidelity movies even when source have poor detail
- general image color tonality (saturation) is often affected (badly)
- in fast motion (long action scenes) is a disaster, especially at low mid-low bitrates (even with good setup in Nandub)
- doesn't encode well non-blurred (but crisp) complex images in low/fast motion sequences
+ best "pixel detail" on very-low motion or no-motion situations

DivX 4.11 (avisynth/mpeg2dll/Virtualdub-YUV):
- tends to smear/approximate (sometimes too aggressively) image particulars with *similar* (pixel) tonality (some areas become flatten)
- conseguences: movies with poor image quality, without a good crispness may get even more negative quality drop
+ manage fast motion more more better than DivX 3.11 (try to encode a pay-TV F1 Gran Prix!! You will see where old DivX3 falls at all!)
+ reflects general image (color) tonality near to perfect
+ *Always* guarantee a superior quality with good crispness/contrast content
~ not perfect but very good "pixel detail" on very-low motion situations


My advise is to use DivX 4.11, but with a good setup like DVD2AVI/avisynth/mpeg2dll (Gordian Knot), and not perform YUV to RGB conversion (set virtualdub to "Fast recompress"), using avisynth engine for filters.

For me Xmpeg is unflexible. (especially in resizing and managing multiple .M2V files)

[pacohaas]

it could be just your video card, i've got a peice of crap video card and it's necessary for me to switch to 640x480 resolution in order for DivX4.x to look good. Make sure you have your monitor set to 24-bit, this is why I have to lower my resolution.

[ChristianHJW]

Originally posted by NeVeRLiFt
I dont see anyone using DivX4.11 making great 1CD rips. They all use Nandub. And I have yet to see a DivX4.11 beat Nandub for making a 1CD rip.

A DivX4.11 1 CD rip will certainly not 'beat' a well done nandub/3.11 encode, but it can give similar quality with a much improved time efficiency talking about the encoding process ..... what i hear from those making very good nandub rips, they will have to invest about a week into a single movie to find a very good profile ... and those rips with 'standard' profiles aren't any better than 4.11 ...

I personally have done a few real nice 1 CD rips using DivX 4, at resolutions of up to 640 * xxx even for 2hrs movies ( normally 560 * xxx to 608 * xxx ) but i admit i had to use heavy filtering ( temp smoother level 2 before resizing, i'm reall happy TheWEV added this feature in GKNot ) to get a clear and nice picture without any macroblocks and they cant compare with the best Nandub rips i came across ...

And for 2CD rips Nandub still has the best quality IMHO

This where i cant agree at all, i'm happy you added the IMHO. DivX4.11 will give 2 CD rips of stunning quality when being used with a suitable encoding method like GKnot 0.21 . And i agree with 'unplugged' that DivX4.11 can handle action/fast motion better ...

[pale]

DivX 4.x handles low bits/(pixel*frame) values much better than 3.11, thus resizing is not a very good idea when encoding with 4.11 (see also http://rilanparty.com/vbb/showthread.php?s=&threadid=10684). The reason why you had to resize with 3.11 was that it handled low bits/(pixel*frame) values poorly so one had to keep bits/(pixel*frame) high by reducing the number of pixels to get enough good quality. Resizing frames of a video is much like resizing any other kind of picture, resizing by itself never improves quality. This is why people inviolved in DTP prefer not to resize, and when this is necessary, then with even dividers (such as 2).

In addition, DivX 4.x produces more blurred image than 3.11 did, so always go at least soft bicubic, preferrably sharp bicubic if you do resize.

[ChristianHJW]

Originally posted by pale DivX 4.x handles low bits/(pixel*frame) values much better than 3.11, thus resizing is not a very good idea when encoding with 4.11

... Hmmm, a 1 CD rip at 720 * 576 , bits/pixel*frame < 0,1 ... very good idea, indeed :D !! Just trying to remember how an anamorphic movie might look without being resized ;) ... you must be a real expert in MPEG4 encoding PALE ... but at least you are right concerning resizing filter ... precise bicubic is the best choice without any doubt for a 2 CD rip at 640 * 272 ...

[PhAtfiSh]

theres no clear difference yet, simple as that. If you cant make divx 4.11 vbr look good then ur an idiot, if you cant make 3.11 look good then you need to learn to use nandub better.

[Peters]

@ChristianHJW

I agree with your comments but in the absolute not resizing is better.And if you go for 2 CD with 640 x 272, no doubt that it's often possible to make a good rip just cropping and using a res like
720 x 304

[pale]

to ChristianHJW:

>>Hmmm, a 1 CD rip at 720 * 576 , bits/pixel*frame < 0,1 ... very good idea, indeed !!

As it seems that I can neither read nor think, could you please explain to me (in plain words please) in which part of my post did I speak about not cropping?

A pretty typical 105 min. movie, 25 fps, cropped to 720*304 with mp3 96-bit audio targeted for one 800MB CD gives 0,175 bits/(pixel*frame). I could be wrong, but to me it seems quite a bit more than 0,1 ;-)

>>Just trying to remember how an anamorphic movie might look without being resized ... you must be a real expert in MPEG4 encoding PALE ...

I admit, that when I spoke about not resizing I should have stated (the obvious) that I meant non-anamorphic sources.

When working with anamorphic sources, try to keep width at 720 (or cropped to 704) and only adjust height. I've done several comparisons by encoding many movies with otherwise same parameters except the resolution and I'am yet to see a movie, where resizing width also would have produced a better result. Although I 'm sure that if I was to make 1CD rip out of a very long movie, that propably would happen.

[ChristianHJW]

Originally posted by pale
to ChristianHJW: A pretty typical 105 min. movie, 25 fps, cropped to 720*304 with mp3 96-bit audio targeted for one 800MB CD gives 0,175 bits/(pixel*frame). I could be wrong, but to me it seems quite a bit more than 0,1 ;-)

O.k., this is clearing things up a lot.

1. You're using 800 MB CDs. I never used them because my burner doesnt support them, and so will most CD drives.

2. PAL 16:9 non anamorphic movies are very very rare

3. 96 kbps audio is no good choice, at least not for a short movie like 100 mins. Even for 2+ hrs movies i always use Lame ABR 112 kbps, better 120 - 130 kbps. Unless you dont care about sound at all ...

4. 0.175 bits/pixel*frame must be encoded using precise bilinear resizing filter ( no way for precise bicubic ) and heavy filtering like temp smoother level 3 - 4. Of course, there are some movies with good compressability so it might work out in the end in a few cases, but its certainly not generally a good thing to do .

5. A 640 * 272 precise bicubic encoding will look better and crisper than a 720 * 304 precise bilinear one in most cases ... and it will play much better on PCs with normal video cards .

Conclusion :

Read Doom9 Guides and follow them to the letter, you will get better DivX encodings in the end, i promise. And he will certainly not recommend to do 1 CD encodings ( he doesnt like them at all for quality reasons ) at full DVD res ....

[pale]

>>You're using 800 MB CDs. I never used them because my burner doesnt support them, and so will most CD drives.

This is true with most burners. Players, however, are different story. Most CD and DVD drives do support 90 min CDR's without problems. There are, unfortunately, execptions. Some drives even play them in win9.x and not in win2k.

>>96 kbps audio is no good choice, at least not for a short movie like 100 mins. Even for 2+ hrs movies i always use Lame ABR 112 kbps, better 120 - 130 kbps. Unless you dont care about sound at all ...

I know this and put it there as an example only. Usually I go for 112 or 128-bit ABR. However, 96-bit mp3 is not that much worse than 112-bit (the difference is there, but it's not that big). When I do want quality audio, then 5.1 ch ac3 is a way to go (unless, of couse, the original audio is DS or DD 2.0).

>>0.175 bits/pixel*frame must be encoded using precise bilinear resizing filter ( no way for precise bicubic ) and heavy filtering like temp smoother level 3 - 4. Of course, there are some movies with good compressability so it might work out in the end in a few cases, but its certainly not generally a good thing to do .

This is where I disagree. Like I said before I tried this with several movies with same parameters (except for resolution) and always keeping the width at 720 (or cropped to 704) was better or at least as good. And yes, I do know that almost everybody recommend to use bilinear instead of precise bicubic for bits/(pixel*frame) values under 0,200. I, however, can not be convinced about this, as in my tests changing mode to precise bicubic produced better quality. I did, however, keep bits/(pixel*frame) values around (and significantly above) 0,160, so with really small values (like 0,140), bilinear with plenty of filter work might (and propably would) produce better quality.

>>A 640 * 272 precise bicubic encoding will look better and crisper than a 720 * 304 precise bilinear one in most cases ... and it will play much better on PCs with normal video cards .

This is propably true. My point is that 720 * 304 precise bicubic is better still than 640 * 272 precise bicubic .


I always use hand set quantisizer values (usually MQ at six or below and mQ 4 or below (set KF's and DF's indivually with LogFile patcher utility), and do 2pass only). I know this is another issue where opinions differ, as some (most, including DivX developers) claim that you should always use default quasntisizer values 12 and 2. Once again, I believe my eyes.

By the way, do you happen to know what are those unpredictable results you could get, if you did not use default quantisizers for two 2pass encoding? The logfile is totally different when I use quantisers of MQ6 and mQ 4 instead of MQ12 and mQ2 (as is the resulting movie). I have encoded nearly a hundred movies with hand set quantisizers and never had any problems with it. So if this is what is meant with unpredictable, I think I love unpredictable ;-)

[Doom9]

here's how I do it... use gknot.. resize to a value that gives you a yellow bits/pixel*frame (don't worry.. you can get away with it).. then use sharp bicubic (divx4 has the tendency to blur much more than divx3.. I would never use bilinear with divx4) resizing.

Going above 640x is a big nono for bits/pixel reasons as well as compatibility with slower computers, and last but not least TDX rules. If you want D1 res.. use the DVD... makes no sense to make a DVD rip at DVD resolutions unless it's a DVD -> DVD-R job.

And.. if even the developers say to not touch the quantizers I'd go by that... want to mess around with quantizers get the divx4 logfile tool from another thread in this very forum.

[ChristianHJW]

pale,

what encoding mode exactly are we talking about ?
Could you pls. very briefly describe your encoding method ?
Do you prefer 1 Pass or 2 Pass ? Thanks ...

[Peters]

Originally posted by Doom9
here's how I do it... use gknot.. resize to a value that gives you a yellow bits/pixel*frame (don't worry.. you can get away with it).. then use sharp bicubic (divx4 has the tendency to blur much more than divx3.. I would never use bilinear with divx4) resizing.

Going above 640x is a big nono for bits/pixel reasons as well as compatibility with slower computers, and last but not least TDX rules. If you want D1 res.. use the DVD... makes no sense to make a DVD rip at DVD resolutions unless it's a DVD -> DVD-R job.

And.. if even the developers say to not touch the quantizers I'd go by that... want to mess around with quantizers get the divx4 logfile tool from another thread in this very forum.

hmmm...with all the respect i have for your work , i disagree (just a little :) )

First, compatibility with slower computers. It seems that you don't care of 'obsolete' sound cards (not accepting 48 khz samplerate)
Why don't you care of 'slower computers' ?
And for these slower computers it will be certainly very hard to play a 640 x 272 movie with AC3 . Do you blame people using AC3?
The choice using 720 (or 704) x 304 is not a choice for the res (not only) It's because the quality is better than using a filter for resizing .
720 x 304 , 110 mn, mp3 160kb . 2 CD of 700 Mo ->Gknot gives a 0.3 bits/pixel*frame . So, where is the pb?

[Emp3r0r]

I consider my computer slightly outdated but my P2 350 plays all my 640x272 content with decoding AC3 sound also with no problems. Now I don't use post filtering but my rips don't need them since they are all 2 CD. The only time I have problems is trying to get my computer to play DVD's without my H+ card. Maybe it is SCSI that makes my computer act like a more powerful computer. Who knows.

PS: ok, i have it overclocked to 392mhz with a FSB of 112mhz I believe. Still not a powerful sytem.

[Peters]

funny because one of my 2 PC (yes i must be rich ) is A PII 350 and not overclocked!
(Dell with ATI rage PRO 8Mo)
Not a recent one, isn't it?
I don't try playing movies with AC3 but it plays fine 720 x 304
movies (of course, not using post proc).

So we have to search olders computers...:)

[pale]

I aogise for quite long a post.

To ChristianHJW:

method in a nutshell:

1)Rip with SmartRipper
2)prepare Vob's with DVD2Avi
3)create Avisynth-script with GordianKnot (cropping, resizing,bitrate calculation, deinterlacing if necessary, filters when I feel they are needed).
4)audio decoding (when not using ac3) with Azid&Lame, Nowadays with BeSweet
5)encoding with DivX 4.11 2pass in Virtual Dub
6)interleaving video & audio in Nandub
7)splitting (when applicable) in Nandub
8)plenty of playing around with subtitles, menus etc. (as you seem to be working with PowerDIvX 4, could you please add support for MicroDVD ini-files? Yes, I know PDivX has its own ini-file.)

In cropping I crop sides to 704 if movie has black bars on the side, if not, leave it at 720. Then I crop all the black bars from bottom and above and to the movie to gain reasonable aspect ratio error (like +/- 0.3). With anamorphic sources I'll usually end up with resolution of 704*304 or 704*384. When resizing I always choose Sharp bicubic, even when bits/(pixel*frame) would be something around 0,175 (which occurs quite often, as I do not resize witdth). I don't often go much under that though, as then I feel it's time to throw in another CD.

After first attemps with DivX 4.x I have used 2pass only (and will continue to do so). While encoding with DivX 4.x 2pass I do not respect the rule of not playing around with "right-hand" parameters for the first pass though. Pretty typical parameters I use are:

Max KF interval 250, MQ 6 (or below), mQ 4 (or below), RCAvg. 200.000, reaction 50, RC U/D 5, FD 0. Bitrate according to GKnots calculator, typically between 800-1200 kbps. These seemed to work for me better (clearer high motion with no visible trade off as worse low motion) than the defaults did (tested this with several movies, of course).

After I tumbled into Dimitry's Logfile patcher, I have pretty much used it almost solely. At first I set credits at different zone and set their parameters (much higher quantisizer values), then I set both Quantisizers for 2 for KF, then fiddling with DF Quantisizers (typically MQ around 6 and mQ 2 or 3) to achieve desired bitrate, then save the log and run second pass with it.

To Doom9:

>>here's how I do it... use gknot.. resize to a value that gives you a yellow bits/pixel*frame (don't worry.. you can get away with it).. then use sharp bicubic (divx4 has the tendency to blur much more than divx3.. I would never use bilinear with divx4) resizing.

My point excactly.

>>And.. if even the developers say to not touch the quantizers I'd go by that...

I don't. Seems that I have a problem about following guides;-) I have seen DivX developers suggest one should leave quantisizers at defaults when using 2pass because they did not work properly. I have tested them with nearly a hundred movies and never had a proplem with them, so they pretty much seem to work with my computer.

>>want to mess around with quantizers get the divx4 logfile tool from another thread in this very forum.

I did, and use it with all my encodings at the moment.

To Doom9 and Peters:

>>Going above 640x is a big nono for bits/pixel reasons as well as compatibility with slower computers, and last but not least TDX rules. If you want D1 res.. use the DVD... makes no sense to make a DVD rip at DVD resolutions unless it's a DVD -> DVD-R job.

This is where I have to disagree with Doom9 (and agree with Peters).
1) hardware wise I have similar results than Peters. Alhough I could care less, if my encodings did not work with older hardware.
2)
>>The choice using 720 (or 704) x 304 is not a choice for the res (not only) It's because the quality is better than using a filter for resizing .
This is excactly what I tried to say. This obviously comes with a trade-off as bits/(pixel*frame) values are lower. As I usually keep bits/(pixel*frame) values around 0,160 or above, I feel that benefits of keeping original width outweight the disadvantages of having to use (slightly) lower bits/(pxel*framne) values.

[everwicked]

method in a nutshell:

1)Rip with SmartRipper
2)prepare Vob's with DVD2Avi
3)create Avisynth-script with GordianKnot (cropping, resizing,bitrate calculation, deinterlacing if necessary, filters when I feel they are needed).
4)audio decoding (when not using ac3) with Azid&Lame, Nowadays with BeSweet
5)encoding with DivX 4.11 2pass in Virtual Dub
6)interleaving video & audio in Nandub
7)splitting (when applicable) in Nandub
8)plenty of playing around with subtitles, menus etc. (as you seem to be working with PowerDIvX 4, could you please add support for MicroDVD ini-files? Yes, I know PDivX has its own ini-file.)

In cropping I crop sides to 704 if movie has black bars on the side, if not, leave it at 720. Then I crop all the black bars from bottom and above and to the movie to gain reasonable aspect ratio error (like +/- 0.3). With anamorphic sources I'll usually end up with resolution of 704*304 or 704*384. When resizing I always choose Sharp bicubic, even when bits/(pixel*frame) would be something around 0,175 (which occurs quite often, as I do not resize witdth). I don't often go much under that though, as then I feel it's time to throw in another CD.

After first attemps with DivX 4.x I have used 2pass only (and will continue to do so). While encoding with DivX 4.x 2pass I do not respect the rule of not playing around with "right-hand" parameters for the first pass though. Pretty typical parameters I use are:

Max KF interval 250, MQ 6 (or below), mQ 4 (or below), RCAvg. 200.000, reaction 50, RC U/D 5, FD 0. Bitrate according to GKnots calculator, typically between 800-1200 kbps. These seemed to work for me better (clearer high motion with no visible trade off as worse low motion) than the defaults did (tested this with several movies, of course).

After I tumbled into Dimitry's Logfile patcher, I have pretty much used it almost solely. At first I set credits at different zone and set their parameters (much higher quantisizer values), then I set both Quantisizers for 2 for KF, then fiddling with DF Quantisizers (typically MQ around 6 and mQ 2 or 3) to achieve desired bitrate, then save the log and run second pass with it.


Pale: you say that in the first pass you set the mQ to 4 or below, MQ to 6 or below and the bitrate to what Gknot says. Dimitry states that both quantizers should be set at 2 and bitrate at 6000kbps for the program to function right.

Can you give an answer to that? Is it ok if you dont follow that?

[pale]

>>Pale: you say that in the first pass you set the mQ to 4 or below, MQ to 6 or below and the bitrate to what Gknot says. Dimitry states that both quantizers should be set at 2 and bitrate at 6000kbps for the program to function right.
>>Can you give an answer to that? Is it ok if you dont follow that?

If I said that, I apologise, that is not what I ment;-(. What I ment is that before I started to use Dimitry's app, I set mQ to 4 or below and MQ to 6 or below and the bitrate to what Gknot says. With Dimitry's app I do as he tells (both Q at 2, bitrate at 6000).

[everwicked]

Cool,
I noticed that when you set the mQ for DF to 2 in divx4log, you get oversized files. Have you noticed something similar?

And my main query is one. As you change the settings of divx4log, it comes up with a video size and a bitrate. What exactly is that video size? The size that u will get when u use these settings? What if you want to use a different bitrate to achieve desired filesize?

I used the formula: divx4log_bitrate*(desired_size/divx4log_size) and it comes up with the Gordian Knot suggested bitrate.
Something cant fit in my mind in all these. If the size of the file is dependant on the bitrate then how come the quality depends on the quantizers?

I guess that i need more info on how the quantizers and the bitrate is related in divx4. If anyone can answer these questions i'd be grateful.

[pale]

to everwicked :

First of all, if you're using any other version of DivX4Log than 0.38, I suggest you get that as it makes everything so much simpler.

The size in DivX4log is the predicted size of the final movie (without audio) in bytes. In my experiments I have achieved very good size predictability (remember that MB is not 1000 KB, though). Usually I end up within 5MB of the size what the utility predicts (with a slight tendency to go 2-3 MB above).

0.38 has an internal bitrate calculator though, that tells you the approximate bitrate that will be the result of the set quantisizers. So instead of defining your desired bitrate and setting quantisizers accordingly, you do excactly vise versa; adjust quantisizers so that the bitrate DivX4Log tells you is as close to bitrate GordianKnot told you as possible.

By setting all the quantisizers (KF and DF) bitrate itself (as set in codec parameters) loses pretty much all of its significance. For example if you set quantisizers in DivX4Log so that the bitrate they produce is say 1000, the size of final movie is almost identical if you set second pass bitrate at 500 or 6000. This is very different if you don't use DivX4Log (or some other app).

As for the quantisizers, yes, it is true that setting quantisizers at 2 increases final avi size when compared with avi encoded with quantisizers, say, at 3. As there are much less keyframes than deltaframes, increasing the values of KF Quantisizer by 1, decreases the size of the avi much less than if you increased the DF quantisizers by 1. Furthermore, as DF's are generated from the KF's, the better (read: less compressed) your KF's are, the better chance you have to achieve a great-looking movie. So I try to keep keyframes at 2.

[j6]

Just wanted to give my impression on DivX3 vs. DivX4 since that what the topic is about. I normally convert to SVCD but just recently started experimenting w/ DivX. I downloaded GKnot .21 and preceeded to encoded Reservior Dogs first using 2pass DivX3 and then using 2pass DivX4 also used soft bicubic resizing for both. Both are one 1CD rip (100 min. movie) and same settings. I played both rips full screen to get a better look at the quality. I must say that DivX3 is better quality overall. Do I need to use different settings for DivX4 than DivX3 so that I can make a better comparison?

[everwicked]

Originally posted by pale
to everwicked :

First of all, if you're using any other version of DivX4Log than 0.38, I suggest you get that as it makes everything so much simpler.

The size in DivX4log is the predicted size of the final movie (without audio) in bytes. In my experiments I have achieved very good size predictability (remember that MB is not 1000 KB, though). Usually I end up within 5MB of the size what the utility predicts (with a slight tendency to go 2-3 MB above).

0.38 has an internal bitrate calculator though, that tells you the approximate bitrate that will be the result of the set quantisizers. So instead of defining your desired bitrate and setting quantisizers accordingly, you do excactly vise versa; adjust quantisizers so that the bitrate DivX4Log tells you is as close to bitrate GordianKnot told you as possible.

By setting all the quantisizers (KF and DF) bitrate itself (as set in codec parameters) loses pretty much all of its significance. For example if you set quantisizers in DivX4Log so that the bitrate they produce is say 1000, the size of final movie is almost identical if you set second pass bitrate at 500 or 6000. This is very different if you don't use DivX4Log (or some other app).

As for the quantisizers, yes, it is true that setting quantisizers at 2 increases final avi size when compared with avi encoded with quantisizers, say, at 3. As there are much less keyframes than deltaframes, increasing the values of KF Quantisizer by 1, decreases the size of the avi much less than if you increased the DF quantisizers by 1. Furthermore, as DF's are generated from the KF's, the better (read: less compressed) your KF's are, the better chance you have to achieve a great-looking movie. So I try to keep keyframes at 2.

I have 0.38 and i am aware that 1mb is 1024kb ;-)

So at the end of the day, what we are trying to do is to set the quantizers in divx4log so that the calculated bitrate appoaches the one in Gordian Knot right?

And if a set gives you 1500+ bitrate what can u do? just change the set?

By the way, what postproccessing lvl do u use?

cheers m8

[pale]

to everwicked:

>>So at the end of the day, what we are trying to do is to set the quantizers in divx4log so that the calculated bitrate appoaches the one in Gordian Knot right? And if a set gives you 1500+ bitrate what can u do? just change the set?

Yes to both.

>>By the way, what postproccessing lvl do u use?

Usually 4. It is default playback setting though, thus it has no influence in encoding quality.

To j6:
>>Both are one 1CD rip (100 min. movie) and same settings. I played both rips full screen to get a better look at the quality. I must say that DivX3 is better quality overall. Do I need to use different settings for DivX4 than DivX3 so that I can make a better comparison?

In my opinion, you should. For starters, use sharp bicubic instead of soft bicubic with DivX 4. If you are not afraid of extra fiddling, try also DivX4Log. And if your source is widescreen movie, you might want to try keeping original width with DivX4.

[NeVeRLiFt]

Originally posted by j6
Just wanted to give my impression on DivX3 vs. DivX4 since that what the topic is about. I normally convert to SVCD but just recently started experimenting w/ DivX. I downloaded GKnot .21 and preceeded to encoded Reservior Dogs first using 2pass DivX3 and then using 2pass DivX4 also used soft bicubic resizing for both. Both are one 1CD rip (100 min. movie) and same settings. I played both rips full screen to get a better look at the quality. I must say that DivX3 is better quality overall. Do I need to use different settings for DivX4 than DivX3 so that I can make a better comparison?

The facts are the facts! Nandub kicks ass and makes the best looking 1CD rips possible. DivX4.xx is new and people are excited and wanna try it, but new is not always better ;)
I have several 1CD rips that I will upload to anyone who wants to see and compare. I'm working on "Gone in 60secs" and almost have it....
its at 772mb at moment. I will have it 700mb and show that a movie over 1hr 30min can be fit to 1CD and still have supra quality.
Long live Nandub!!!!!!!
:D

[ChristianHJW]

One thing about your 'perfect' DivX3 rips :

If DivX Networks decides to stop support of DivX3 in their DirectShowFilter one day you'll have to reencode all your movies .... because from my point of view there will be nobody able to port divx32.ax to next generations OS ;) ...

[omol]

Originally posted by ChristianHJW
One thing about your 'perfect' DivX3 rips :

If DivX Networks decides to stop support of DivX3 in their DirectShowFilter one day you'll have to reencode all your movies .... because from my point of view there will be nobody able to port divx32.ax to next generations OS ;) ...

There's always Linux come to rescue.....;)

regards
omol

[Peters]

Originally posted by ChristianHJW
One thing about your 'perfect' DivX3 rips :

If DivX Networks decides to stop support of DivX3 in their DirectShowFilter one day you'll have to reencode all your movies .... because from my point of view there will be nobody able to port divx32.ax to next generations OS ;) ...

Since it works very well with XP, you have time, don't worry..
And if Divx Networks stops any activity, you'll have to reencode all your Divx4 movies ? :)

[everwicked]

Originally posted by NeVeRLiFt


The facts are the facts! Nandub kicks ass and makes the best looking 1CD rips possible.


Ofcourse it does! :o

[ChristianHJW]

Originally posted by Peters
Since it works very well with XP, you have time, don't worry..

... its probably due to my age that i am thinking in a longer term than you do .... ;) ..... what's the average life cycle of a M$ OS ??

And if Divx Networks stops any activity, you'll have to reencode all your Divx4 movies ? :)

..... by having a little background about DivXNetworks i would tend to say that this is very very unlikely ... and if they do you can be sure that guys like GEJ, Junto or Sparky will make sure the code is alive somewhere ;) ...

[ChristianHJW]

Originally posted by NeVeRLiFt
The facts are the facts! Nandub kicks ass and makes the best looking 1CD rips possible. :D

Agreed ! But only for now .... wait for DivX 4.5 .... and DivX 4 specific tools like BlackSun's 'Morphix' or Dmitry's tools are yet to get effective with this version as it will finally allow log file editing ....

[NeVeRLiFt]

Originally posted by everwicked


Ofcourse it does! :o


This guy everwicked is a dumbass and really talks alot shit about Nandub and he does not even know how to use it.




:cool:
Stupid ass!
http://www.rage3d.com/board/showthread.php?s=&threadid=33594151

[everwicked]

Noone will bother to answer your stupid post like the previous one. I can't understand why i do...:(

[NeVeRLiFt]

Originally posted by everwicked
Noone will bother to answer your stupid post like the previous one. I can't understand why i do...:(

Hey stupid why dont you learn how to use a software(Nandub) before bad mouthing it. Cause you really look the fool talking shit about something you dont nothing about.
:cool:

[NeVeRLiFt]

Yo Doom9 did you not say Nandub was best? Doom9 did you not say the quality and size Nandub can do makes it King? These n00bies run out here using DivX4.xx since its so easy to use and think its be all end all for DivX! Well there fucking wrong and if they would take the time to learn how to use Nandub they would see real quality.

NeVeRLiFt

btw everwicked lets see how everstupid you can get trying to get out of this one ROFLMAO ;) you know Doom9 and Nicky got the guides... can you not follow them and see for yourself? Maybe you need help setting Nandub up so you can get good quality with it? Just because you dont understand something or know how to use it does not mean its bad and you should fear it.

[everwicked]

Results will talk.

Get your DSL running and shut up.

[Doom9]

did I say nandub was better than divx4? don't think so http://www.doom9.org/codec-comparison-update.htm

it's a case by case thing.. sometimes you can get better results using nandub, sometimes divx4 is better.

and since this thread has degraded into a flamewar it's time to close it.