matrix revolutions ugly xvid encode

[chrome]

im trying to rip my revolutions dvd to xvid, but every time i try, it ends up looking relatively ugly, compared to my other rips... im aiming for 2 cd rip, 608x256 resolution, with ac3 sound... has anyone ripped this successfully? if so can you share exactly what settings you used, and maybe a screenshot of your encode? does this film just not rip nicely or something? any help greatly appreciated. ive tried lots of different settings, but im just not happy with the results im getting. thanks.

[Neo Neko]

Um no. It is not released till the day after today. [3:21 Monday morning] How did you get it?

[chrome]

i think its only released in the states in 2 days. here it was out on friday. so i guess ill wait till you guys get it. cheers anyway.

fyi, the dvd was pretty rubbishy, only like 2 good bonus documentaries worth seeing... ugly ass cover, too!! guess we have to wait for matrix revisited 2 or something, for the good stuff...

[Mug Funky]

i'm racking my memory of the film to see if there's any particularly encoder-challenging scenes in it. can't remember any. if anything i'd think it an easier encode than reloaded (the lighting in that movie is inexcusably ugly and amateurish, considering the great lighting in the first and maybe the last movies)

[Teegedeck]

[A discussion of the quality of the original Matrix Revolutions DVD and other DVDs has been split from this thread. Please continue this discussion there (http://forum.doom9.org/showthread.php?&threadid=73966) .]

[Tuning]

Matrix3 has many codec killer scenes like, Neo-Smith fight in background of lightning, Sentinels moving like chain in background of fire (I mean explosion)etc etc...

Therefore I think it will be more difficult than reloaded. ;)

[Soulhunter]

Originally posted by Tuning
Matrix3 has many codec killer scenes like, Neo-Smith fight in background of lightning...Dont forget the rain... :D


Bye

[Tuning]

Originally posted by Soulhunter
Dont forget the rain... :D
Most often lightning is preceded or followed by rain!:D

[chrome]

while this is a facinating discussion... it doesn't really address the original question... how do i make a nice f00kin rip of this sucker!!!!

[Soulhunter]

Originally posted by Tuning
Most often lightning is preceded or followed by rain!:D Also the ones Neo shoots out of his hands ??? :D :D :D

chrome
while this is a facinating discussion... it doesn't really address the original question... how do i make a nice f00kin rip of this sucker!!!! Lower resolution, de-noising and/or 3CD's maybe ???


Bye

[chrome]

how important do you think the max overflow improvement / degradation settings on the second pass, will be in this case... what values would you suggest/use?

[Tuning]

@chrome

What are the XviD settings in use?:)

[Blue_MiSfit]

yeah i had to do a really high resolution and bitrate encode with andreas 78er to get it looking nice. I think it was at 1024x something, and the settings were something like this:

2 pass - 2gb target size
MSP @ 6
VHQ @ 4
B-Frames: 5/1.5/1.0
Qpel
GMC

My AVS was pretty simple - just UnDot if I remember correctly. It was under RC3 with VDubMod 1.5.10. Also, I had the vdub sharpen filter set to 10 or 15 to help with the (lanczos) resize.

I just muxed in the 6ch AC3 and got a Matroska file about 2.5 gigs - looked real nice, but was crazy slow with full postprocessing with ffdshow. With light PP it was nice, even better with a slight level correction.

Ultimately I decided it was too big to be practical, and re-encoded it at a much lower bitrate at 960x something with less ASP settings and hvs-best. Not so nice, I am figuring out how to get it looking good at more reasonable bitrates.

good luck anyone trying to encode this shit. for comparison, I got return of the king to look nice at 2 gb, and its something along the lines of 2 times as long. bahhh..

-misfit

[Koepi]

If you use vdub filters, the video will be converted to RGB and in xvid back to yv12. Why dopn't you use a real YV12 process chain which will give the "best possible input quality"(->fast recompress in vdub)?

Koepi

[Blue_MiSfit]

I did not realize this kopei - I will be sure to use all avisynth filters in the future ;)

on another note, just for fun today I started an encode using Xmpeg, with no crop or resize (4:3 version), with the MPEG quants, qpel, gmc, and bframes at 3/1.5/1. I also enabled trellis. as usual msp @ 6 and vhq @ 4.

So it was a full resolution encode at approx 1350 kbit. I turned on the output preview window and things were looking very very nice - even in the extremely high motion first battle between the APU corps and the sentinels (amazing).

We'll see in the motning
~misfit

[Koepi]

Matrix Revolutions, PAL DVD (region 2).

AVS script:

LoadPlugin("h:\Software\avs-yv12\MPEG2Dec3dg.dll")
LoadPlugin("h:\Software\avs-yv12\UnDot.dll")
LoadPlugin("h:\Software\avs-yv12\SimpleResize.dll")
MPEG2Source("h:\Filme\matrix3\matrix3.d2v",CPU=0,idct=7)
LumaFilter()
Crop(0,76,720,424)
UnDot()
Trim(0,172862).SimpleResize(640,272)+Trim(172862,0).BilinearResize(640,272)


Encoding using vdubmod 1.5.4.1, all xvid settings standard except:
- adaptive quantization
- qpel
- 3 bframes
- hvs better picture quantisation matrix
- trellis quantisation
- vhq4
- Zone starting at 172862 fixed quant 16, begin with keyframe checked, bvop sensitivity 50.

Desired size 620804, real size 620562 (2 audio streams).

The result looks quite pleasing, sometimes very little blockyness shows up in fog. Using postprocessing makes that invisible though. Details are good.

But speaking generally, this movie is a codec killer again. As you can see i needed to use some tweaks like SimpleIDCT for decoding the mpeg2 and use SimpleResize instead lanczos.

Regards
Koepi

[Jawor]

use SimpleResize instead lanczos
I suppose we can use Lanczos at 2 CDs.

Here's my script for 1 CD (Polish edition, Regions 2 & 5)
MPEG2Source("...")
Crop(2,76,-2,-76)
Deen("c3d",1,6,8)
UnDot()
BicubicResize(624,256,0,0.5)
UnDot()
I followed ITU-R BT.601, hence the resolution. I cropped less than I should according to GK, because the image was too flat after cropping 6 pixels left and 6 pixels right. The desired size was 629 MB (I used Ogg Vorbis audio at quality=2.00, English soundtrack, 73 MB). I got perfect 702 MB in a Matroska file.

The result doesn't look brilliant, but it's quite good. The avg. quantizer is about 4.5. According to StatsReader, the first pass size is ca. 1400 MB, so I suppose that this script (with LanczosResize instead of BicubicResize) could generate a great 2 CDs encode of this movie.

Here are the XviD (1.0 RC3) settings:
AS@L5
quantizers 2-31
B-frames 3/1.50/1.00
Trellis, Qpel, AQ & GMC ON
Packed Bitstream OFF
HVS-Good Picture Quant Matrix
MSP=6, Chroma Motion ON, I-frame interval 250, VHQ=4
B-VOP Sensitivity 1 (movie), 6 (credits)
Weight Zone = 0.20 for the end credits
I left the rest at default settings

I suppose my encode was that successfull thanks to Deen()+2*UnDot() :D

[Sagittaire]

A good movie for 1CD Codec Comparaison ... certainely better than first Matrix ... ;-)

[chrome]

just to let you know, youve helped me out quite a bit... with each encode, the quality im getting is getting better every time i try something new... about to set another one going overnight, so ill see how much closer i am in the morning... but anyway, thanx for the help guys.

[chrome]

ok, heres where im at, at the moment. personally im pretty happy with it... i took some pics of the more uglier scenes (that is, scenes i thought would be among the more taxing ones, i.e. lots of motion, very dark sequences, or ones a place i found my original tries to look really nasty.) i think m6.jpg is the ugliest of the lot, but its quite an improvement over my original attempts.

anyway, here are the links (ill take em down in a day or so):

http://www.klauke.net/m3/m1.jpg
http://www.klauke.net/m3/m2.jpg
http://www.klauke.net/m3/m3.jpg
http://www.klauke.net/m3/m4.jpg
http://www.klauke.net/m3/m5.jpg
http://www.klauke.net/m3/m6.jpg

this is for a 2 cd, ac3 rip, btw, at 608x256. feedback would be nice. saying its an ugly rip, wouldn't. :) cheers to all for any and all help youve given, though.

[bond]

Originally posted by chrome
feedback would be nice.hm for feedback it would be great if you would post your encoding settings (avs script) ;)

also maybe providing screenshots from the dvd to be able to make frame-by-frame compare would be nice :)

[kurt]

Originally posted by Koepi

...
Encoding using vdubmod 1.5.4.1, all xvid settings standard except:
- adaptive quantization
- qpel
- 3 bframes
- hvs better picture quantisation matrix
- trellis quantisation
- vhq4
- Zone starting at 172862 fixed quant 16, begin with keyframe checked, bvop sensitivity 50.

Desired size 620804, real size 620562 (2 audio streams).
...
Koepi [/B]

Hi Koepi,

why do u use vhq4? I thought, that vhq1 would made better results...and what about the bvop sensivity? I used in all of my rips 0.

[udjathi]

Well, i've already backup Matrix Revolutions, it's quite good, and only last 2 hours, so i went for one cd, i'm using Xmpeg, that works perfectly for me..

Settings

663 Kbps
Trellis
B-Frames default
Adaptative Compression
GMC
624x256 (mod 16)
I-Frames 2-10
B-Frames 2-31
P-Frames 2-10

There is some artifacts in the figthing sequence with Smith, but they are minor

[Teegedeck]

Originally posted by kurt
Hi Koepi,

why do u use vhq4? I thought, that vhq1 would made better results...and what about the bvop sensivity? I used in all of my rips 0.

VHQ=1? This is very old information.

For ages now it's been known that you should set VHQ as high as possible if you want quality to be as good as possible.

B-VOP sensitivity=0 is correct. Koepi wrote that you should use a higher sensitivity for the credits.

[kurt]

oops, sorry - my fault.. :D
but where can i get this: "- hvs better picture quantisation matrix" ?

[Teegedeck]

It comes with Koepi's XviD binary. Look for an archive with matrices in your XviD directory on your HD.

[Sycho]

I've notice that the PQ of revolutions is not that great, as the shitloads of artifacts throughout the film, could that have something to do with it?

[jjseth]

while this is a facinating discussion... it doesn't really address the original question... how do i make a nice f00kin rip of this sucker!!!! Try NewAVSoptimizer in ARCalculator. Incredible results!!!:D

[chrome]

Originally posted by jjseth
Try NewAVSoptimizer in ARCalculator. Incredible results!!!:D

downloaded, and installed, and in the middle of muddling my way through it... would you be able to write, or post the link too, a walkthrough on how to use it properly? all the help is written in spanish, but from the sounds of what this program does, it really sounds quite interesting.

[jjseth]

Help file was written by me but my english...:( Maybe Mashabak, DarkDudae, etc can help you. In this thread (http://forum.doom9.org/showthread.php?s=&threadid=62286&highlight=ARCalculator) you can find a correct explication.
NewAVSoptimizer analyze and detect the scene's complexity to apply a different filter in each case. Results of 50% in compressibility test are the perfect opportunity of make a .avs optimized to gain the 65% target that provides a good encoding reducing the possible artifacts.
NewAVSoptimized make a .avs optimized with every case filtered.
Select the .d2v and make crop, credits selection, size, finally aspect ratio (make compressibility test before), etc. Finally click "calculate". Go to File menu and select "Save as .avs".
Choose LumaFilter, Undot, and LanczosResize and press NewAVSoptimizer.
Select a profile (default is a good choice) and run NewAVSoptimizer (you can run the analysis over a 30% of the frames for example like the percentage of compressibility test but the 100% make a best analysis). Configure XviD and wait (is like a ahead third pass).
Then you can encode the film with the .avs optimized created.

Sorry my poor english.

[chrome]

your english is fine, i think i understand most of what you said. you said do a compresability check ... that is found in the "save as .avs" option, right? its a button down the bottom... or is there another way of doing it (it sounds like you said i should do this, even before touching anything in the "save .avs" options.

also, what will compressability check do for me? is it info i can use, or info that the PROGRAM will use to help the encode look better?

once i have the optimised avs, i can continue to use the program to do the encode, or should i use vdub, or wont it matter? or is it better and easier if i use this program to do the final encode?

thanx for your help. i guessed most of what i had to do in my initial go, but its nice to have some confirmation from someone who knows what they are doing.

-edit: another question, when i go to click the new avsoptimiser option, i put it on 100%, then i have to configure the xvid codec... as i will ultimately use the avs file to do a 2 pass encode, for this pass, do i just configure my 1st pass, then click ok? hope you understand what i mean.

[jjseth]

After compressibility test you'll be redirected to principal window (Date/Datos tab) and you can change the resolution to find a optimun percentage. After configure resolution/compressibility test porcentage select again "save as .avs" and run NewAVSoptimizer.

Compressibility check beside to protect you of undersized problems give you the optimun porcentage to apply the NewAVSoptimize. A result of 50% plus the optimized filters can give an optimun 65% target in XviD encoding.

The optimized .avs can be used with VDub, VDubMod, etc, but ARCalculator include and option call AVS2avi inside File menu. AVS2Avi is a excellent light and fast encoder, i recommend it you.

When you go to run NewAVSoptimizer put it 100% and configure the XviD like a first pass, then clic ok, choose the api versión of XviD (normally the Dev-api4=XviD 1.0) and run the application.

I hope help you...

[chrome]

yep that all pretty much makes sense to me, and was what id allready done before checking back here. im in the middle of doing the new avsoptimiser bit, and ill let it encode overnight, so ill see how that turns out for me. if i have any more problems, ill post again here. thankyou for your help.

oh, another thing, with setting the end credit frames in the program: i read in that thread you posted before that this option doesn't work properly yet. should i be setting these in the program, or just do it in the codec settings?

[jjseth]

With ARCalculator 0.9.4 credits configuration works properly; you can configure credits zone from the program without problems, but if you want you can changed this configuration in the codec. Default configuration quant=18 and greyscale for black and white credits is a good choice. NewAVSoptimizer apply the necessary filters in that frames.;)

[Enigmax]

The results with AR Calculator are fantastics.

Great program.

Greetings:)