Extra Detail Profile Public Test 1.3

[CruNcher]

Extra Detail Profile Public Test 1.4.5

changelog
+ SSE2 Code sad16/dev16/fdct only used in the Intel build
+ removed Best Profile
+ two builds now AMD and Intel
+ synced with CVS Head 1.2

Extra Detail Profile Public Test 1.4

changelog:
+ changed to Foxers RC
+ added 3 unrestricted Main Profiles (Good,Better,Best)
+ Skals updated Fdct/Idct (SSE2)
+ changed to a unified build (still needs testing)
+ cartoon mode save now

There 2 unrestricted (non vbv) Profiles that can be used now
Good (Fastest) = The fastest should be used all the time Speed/Quality Balanced (should work on every Standalone Generation)
Better (Fast) = Even more Details are Preserved (useing Qpel) (should work on 2nd Generation Standalones)

Download here:

AMD Build (http://cruncher.mufflastig.com/XviD/test/XviD-Extra_Detail-1.4.5-AMD.7z)
Intel Build (http://cruncher.mufflastig.com/XviD/test/XviD-Extra_Detail-1.4.5-Intel.7z)

All your feedback is welcome
source,runtime(min),resolution,profile,chosen bitrate or size,final bitrate or final size,1st pass encoding time,2nd pass encoding time,processor speed in real Mhz/Ghz and core

Example = 2010DVD,116,704x324,Good,1000,998,00:52h,01:17h,AMD 64 2Ghz Venice

Standalones tested so far

Medion 7475 (MTZ firmware) (MT1389GE) (Good,Better) (Anamorph compliant)
Siemssen SCO 5000 ND (2nd Firmware) (Vadis 776) (Good,Better) (.avi non Anamorph compliant)

Q: What is it optimized for ?

A: The Extra_Detail Profile (settings) used here were optimized tested and tweaked for
Cinemascope 2.35:1 2.40:1 and 1.85:1 DVD Movie resolutions upto 720x576 D1.

Q: Down to which bitrate does it still gives good results ?

A: This depends on the Source as we all know, but it should do really well
with some hard DVDs for 720x304 2.35:1 that have a runtime of arround 3 h @ 600 kbps.
Easy formular to describe it: The more clear (less noisy) the DVD the better results
it gives and the less ringing will occour @ low bitrates.

Q: What is the main purpose of this ?

A: The main purpose is to create situation specific Profiles for a yet to be finished new
XviD Gui also Frontends as well as Detail Preservation considerations.

Q: Which Matrix did you use here ?

A: I came to the conclusion for now that HVS-Better is the "Best" Matrix for this use
maybe thats gonna change in the future.

Q: Which PSNR did you tried to achive in your tests ?

A: 2.35:1 720x420 Low Bitrate (600) = 40-43 dB, High Bitrate (>700) = 41-44 dB
2.40:1 720x420 Low Bitrate (600) = 40-43 dB, High Bitrate (>700) = 41-44 dB

But Visual Tests where given more priority then PSNR or SSIM results :)

Q: How to install this there is no installer ?

A: Yeah i know i wanted to make this as small as possible (without useing upx) so i decided
for no Installer @ the moment if you allready have a XviD 1.0 beta or rc installed or the
final 1.0 you can just backup and overwrite the files in your %windir%/system32 folder that
are in the *.7z achive except the *.inf file that you can also use to install this just
right click install :)

Q: On which build this is based ?

A: CVS HEAD 1.2

Q: How fast this will be encoding usually ?

A: Jesus that's really scene,filters used and bitrate dependent, but it should be like 10-18
fps with a P4 2.1 @ 720x420

Q: How should i use it ?

A: First LOAD DEFAULTS ;) Select one of the Extra Detail Profiles (Good,Better,Best) encode
1stpass after that change to 2npass chose bitrate or size and do the final encode.

Q: Which is the recomended way for Encoding with this ?

A: The Manual way ;) Cmd + Notepad + Avisynth + Avs2Avi and Virtualdub were used for all Encodes made with
this Gknot or AutoGknot were not tested it should work with them tough, but please don't
change the Settings with the internal Gknot/AGknot configuration that are not enabled in this
build and post that as valid Results :(
This also counts for every other Frontend thats available (avi.NET)

Q: Which Setting you suggest to change ?
A: Just the Maximum I-Frame Interval in the Advanced Options Tab :)
240 FILM
250 PAL
300 NTSC

Q: Will this work for 512x384 or higher 4:3 Res Movies ?

A: Yes but i can't say for sure as i didn't tested with those resolutions

And a big big thx to all the Beta Tester (superdump,insolit,dcoder,spyder482,bonzi,deathwolf,manao)
syskin and gomgom off course for sustaining all my reports same goes for prunedtree and everyone else
in #XviD and #mpeg i hope i didn´t forget someone.

[RadicalEd]

http://misatokatsuragi.home.comcast.net/vde.gif
On configure.
:*|
Athlon tbird, if that's of any use...

[CruNcher]

do'h sh*t :rolleyes:
wait whats a Tbird for a config ?
no SSE right ? hmm it should at least work with XPs
@RadicalEd is only the vfw interface crashing what happens if you try to encode ? does the same also happen with rundll32 xvidvfw.dll,Configure ?
Any Amd users for whom it's working ?

[RadicalEd]

Right, no SSE. Also, trying to configure directly is a no-go. As for trying to encode...
http://misatokatsuragi.home.comcast.net/vde2.gif
:*/

Is it in CVS anywhere?

[Seiby]

Originally posted by CruNcher
Any Amd users for whom it's working ?

Works fine on my XP 2000+, but it seems slower to me than the "normal" xvid 1.0...

[CruNcher]

@ RadicalEd nope not in CVS
please try this http://cruncher.mufflastig.com/XviD/test/XviD_Profile-Test-Tbird.rar

@Seiby
Yes that's normal read about the speed in the Q&A above :)

[RadicalEd]

Sorry to report it, but the results are the same :\

[CruNcher]

@RadicalEd
please download again the Tbird version this should work now :)

[RadicalEd]

Yes! That did it. Awesome, thanks. :D

[CruNcher]

No comments everyone Happy ?

[Sharktooth]

Im sorry but im trying the celticdruid 1.0+trellisfix build.
As i finish testing it i will move to your build.

[JohnMK]

I'd use it if it were more flexible . . . . thanks for the effort but it's just not for me.

[CruNcher]

Slowly i beginn to fear that nobody is interested in that topic more into other ehm useless things well ok your choice.

[JohnMK]

I'm not sure I understand that, did you have some greater purpose here or were you simply trying to be kind & provide a build with the trellis fix?

[Soulhunter]

Originally posted by CruNcher

Slowly i beginn to fear that nobody is interested in that topic more into other ehm useless things well ok your choice.
Ok, after finishing some unless things, I will download it... ;D


Bye

[Seiby]

Originally posted by CruNcher
Slowly i beginn to fear that nobody is interested in that topic more into other ehm useless things well ok your choice.

Well, I've done a small comparison between Extra Detail and XviD 1.0 with my favorite settings (Custom MPEG Matrix, Qpel) on the movie Battle Royal 2 with a resolution of 672*368. Desired bitrate was around 1850 kbits (1690 MB) for a 3CD encode.
I'm not a perfect quality analyzer, but there are definitely more details in high motion as well as in low motion scenes. The encode was only around 5% slower, but the boost in quality is worth this extra time.

Seiby

(Sorry for my crappy english, still learning ;) )

[Soulhunter]

I see, its some sort of "pre-configured for low bitrates" version !!!

Really n00b friendly build I think... :)


Bye

[CruNcher]

*updated*

[Sharktooth]

Uhm... since xvid supports profiles can't you just compile a standard CVS build but with your "custom" profile(s) added to the list?

[PlazzTT]

Thanks for the update Cruncher.

Last week I tried to get The Last Samurai (2hr 28mins) on to one CD. AutoGK used a bitrate of 572 and the quality wasn't great.

I'm encoding the same AVS now with your profile, it should be a good test for it.

[CruNcher]

@Sharktooth

Yes indeed i will look into this posibility, but this is more a wide range test and i used this way of setting since i started when no custom profile existed so im familar with that and as i said it's important to minimize user errors to get the most adequate results this way.

@PlazzTT
Thx every Test and feedback is appreciated

here is one of my test results with Deep Blue Sea (complex) 100m

http://cruncher.mufflastig.com/XviD/extra_detail_standard.png

the red curve shows the distibution with the Standard RC settings
and the blue curve shows my tweaked settings

[BoNz1]

Hi, CruNcher. I tested your tweaked XviD on LOTR ROTK. It seems to have got reasonably good results. I know you want to keep everything simple and all. But what exactly did you change with the rate control? Could you post a patch or even just the source file? I hate having all this config hidden although I realize this is sort of the point of this binary.

Also, I think that having adaptive quantization is having a bad effect on the quality. In ffdshow if you turn on the quantizer display it seem to me that when adaptive quantization is active the quality is not consistent throughout the frame. Seems to me that the goal of adaptive quantization should be to make quality more consistent throughout the frame.

I was recently reading a paper on this that detailed an improvement of Zhihai He's rho domain rate control for b-frames and some other instances where the first rate control had a lot of problems. In addition to the rate control it presented a algorithm to do adaptive quantization using similar techniques. It might work really nicely for XviD. If anyone is interesting in implementing this here is a link (http://www.ece.gatech.edu/research/labs/MCCL/pubs/dwnlds/kim_icip04_2.pdf) . I guess it would be possible to just do adaptive quantization as described by this paper but I think it would be sort of a shame not also do the rate control. I think if your tests show us anything it is that there is still a lot of improvement that can be made to the rate control.

[Teegedeck]

Originally posted by BoNz1
In ffdshow if you turn on the quantizer display it seem to me that when adaptive quantization is active the quality is not consistent throughout the frame.
BoNz1, are you really talking about quantizers (because that is what ffdshow shows) or about quality? Assigning quantizers on a macroblock base, not per frame, is exactly what adaptive quantization is about. The aim is to retain a consistent perceived quality in a frame while bringing the filesize down. Personally, I find it reaches that goal. The savings are translated into an overall quality-gain in second pass.

If you compare two-pass encodes with and without adaptive quantization, don't you find the overall effect beneficient?

[BoNz1]

Originally posted by Teegedeck
BoNz1, are you really talking about quantizers (because that is what ffdshow shows) or about quality? Assigning quantizers on a macroblock base, not per frame, is exactly what adaptive quantization is about. The aim is to retain a consistent perceived quality in a frame while bringing the filesize down. Personally, I find it reaches that goal.

Yes, I know what it does. You can see this if you go to visualizations and enable quantizers. And I am talking about quality. What I am watching in ffdshow is quantizers in every macroblock, when they are different throughout a frame then I know that AQ is active. Then I try to judge the consistency of quality throughout the frame.


The savings are translated into an overall quality-gain in second pass.

If you compare two-pass encodes with and without adaptive quantization, don't you find the overall effect beneficient?

No, to be honest if you look at the quantizers in ffdshow adaptive quantization is only used very rarely. Maybe I'm being a little picky since it is easy to find frames were the effect it bad. It is used so rarely I doubt there really is any benefit so maybe it is safer to leave it off. Anyway, I don't want to take this thread off topic. Back to testing everyone! :)

[Teegedeck]

The 'benefit' was about 10% bitrate-savings when I last checked; it really is a worthwhile feature.

[BoNz1]

Originally posted by Teegedeck
The 'benefit' was about 10% bitrate-savings when I last checked; it really is a worthwhile feature.

Nope, I'm not sure what kind of video you are testing this on. If I do a fixed quant encode, the bitrate reduction is approximately 3% sometimes even less. I tested 9000 frames LOTR in a scene where AQ was used most. The file size without AQ is 82,545KB and with AQ is 79,716KB. Certainly not that big of a difference and I have seen enough frames where the quality is too inconsistent throughout the frame for me to use it.

[Teegedeck]

Yes, my friend, you seem to be right. About 3% on some trailers I just encoded. My apologies.

[Sharktooth]

With "hard" sources even 3% is something worth taking into consideration...

@Cruncher: Using profiles you may also add profiles for high, medium and low compression.
It will be a totally custom xvid release... not a bad idea :)

[PlazzTT]

A bit about Cruncher's build and a link to this thread has been posted in the news on the front page of Warp2search.net

Is this codec really ready for public use already? :confused:

[CruNcher]

it should be stable as XviD 1.1 is @ the moment, but indeed wide spread distro is a little to early
i knew that this could happen hopefully at least some decide to join up and give some feedback btw feedback how was your last samurai encode ?

[Sharktooth]

The results are not bad at all. But...
I tested with 3 differet "average" bitrates: 800kpbs, 1100kbps, 2000kbps @ 720x400 (anamorphic).
The last 2 look pretty good but the 800kbps encode has some "bad" frames. I mean the quant in those frames is too high.
There are also some extraordinary looking frames that however look good (but not so much) in the standard 1.0.3 encode too. Seems you have modified the curve scaling in a way it allocates more bit when it is unnecessary and less bits where it needs them...
However the "average" quality seems to be better than xvid 1.0.3 standard settings (+trellis+vhq4).

EDIT: i will encode the clips with 1.1 from CVS too.

[PlazzTT]

My Last Samurai encode: I forgot to use the same AVS for both encodes. But I'm redoing both now. (ie- letting AutoGK do it's encode automatically with XviD 1.1, and then encoding AutoGK's AVS file with the Extra Detail Profile 1.3. (what's the shorthand name for your codec, btw? EDP? :) )

[PlazzTT]

Some screenshots of EDP 1.3 (Cruncher) vs CVS 1.1:

http://www.eirways.com/lastsamtest/

I used 1 B frame, no QpeL, no Adaptive Quant, no GMC, Chroma Optimizer enabled, VHQ 1 in both.

I have the original AVS, the CVS 1.1 encode and EDP 1.3 encode here.

How can I run a PSNR or similar to compare them (without having to do a re-encode)?

[CruNcher]

use this script replace b with the content of your original avs, compareyv12 can be found on the forum or Wilberts avisynth plugin list. Run this avs then and wait the way i do it is in mpc via cmd
mplayer comp.avs /minimized /play /close

name="name_of_the_XviD_encode.avi"
a=avisource(name).trim(1,0).assumefps(100)
b=MPEG2Source("t3.d2v").crop(0,78,720,420).DRemoveGrain(mode=8).assumefps(100)
compareyv12(a,b, "YUV", name+".psnr.log", false)

[Sagittaire]

http://cruncher.mufflastig.com/XviD/extra_detail_standard.png

IMO generaly for MPEG4 ASP without PP (espacially deblocking) Red line RC is better than blue line RC : perhabs better Average PSNR for blue line (better quality in low motion) but better Overall PSNR for red line (better quality in high motion ... quality more constant).

Lower bitrate modulation (higher quant for high motion) is a good RC strategie for MPEG4 AVC (inloop with high motion -> visual quality degradation is not very important for MPEG AVC) but not for MPEG4 ASP ...

Could be a good RC strategie for short and very high motion part (for exemple low motion : 90% of frame and 50% of size and high motion : 10% of frame and 50% of size).

[CruNcher]

@Sagittaire
From the HVS part you're wrong you shouldn't really concentrate to much on PSNR values, as i stated i checked the degradation in High Motion parts and tried to hold the quality lose as small as possible and i will go on tuneing this based on HVS points, sure the duration of the High Motion part is an important isue the longer it lasts the more visible the degradation gets for the viewer and thats one of the things i will get on soon.

[Gannjunior]

Ciao Cruncher,
i've seen in Sharktooth test the really impressive result of xvid EDP and I want to try it. My cpu is athlon XP (core Barton) so I've downloaded the SSE.rar : i've backupped xvidcore.dll and xvidvfw.dll and I put your dll: however i think something went wrong: reading your FAQ i unserstand it should be possible to change,for example, custom matrix (as u suggest to use hvs-better..), but I've lost differents thing in the interface: the possibility to change matrix, chroma motion, qpel etc...if i come back with old dll everything goes right. Have u got any idea? Is that normal? what i'm wrong?
ah, i just loaded default settings as first thing.
...i'm really impatient to test your EDP! ;)

TIA

[CruNcher]

@Gannjunior

no everything is ok those things are presets so you can only change the things that are visible :)

[Andrey]

The hardest to encode movie I own is Star Wars II.
Last time I've tried to encode it to one CD with XViD 1.0 branch I failed miserably - ringing killed that encode.
Ok, I'll give this build a try and will try to compare it with Nero h264 encode...
I hope results will be ready till next week ends (our week starting at mondays :))

[ookzDVD]

I've got oversize problem while encoding short trailer clips,
about 2 minutes.

[Sharktooth]

With short clips the ratecontrol may fail.
If you get an oversize try to limit the quants to 2-31.

[Didée]

Originally posted by Sharktooth
If you get an oversize try to limit the quants to 2-31.
Exactly. Only I would make that sentence shorter:

"Limit the quants to 2-31."

;)

[Andrey]

Hi guys !
Sorry, I'm too late, (my good friend died, I will be off for some time, I think).
Anyway, my results on Star Wars 2:
Not bad, better that with my own settings and h263 quant.
I think the matrix did most of the job...
Movie looks great when playing on original resolution, but while stretched to the full screen sometimes it looks very ugly :))
Ringing still present, but not as strong, as it was before.
1. Ringing (or whatever it is on the edge :) ) example:
example 1 (http://sirgrey.sbn.bz/videoTest/ex.avi)
See it on top of the officer head - some moving macroblocks and trail after officer get off. Such effects present everywhere thru the film, but original dvd do not contain them. With h263 the effect is much stronger, forming contrast line around edges sometimes.
2. Myst.
See for yourself.
example 2 (http://sirgrey.sbn.bz/videoTest/ex1.avi)
example 3 (http://sirgrey.sbn.bz/videoTest/ex3.avi)
Scene wuth ship gates have fussy macroblocks because of moving myst, I think. This film have a pretty huge quantity of such a scenes: rain, flying sand and so on.

[CruNcher]

@Andrey
Sorry to hear that with your Friend my deepest condolences to you and his family :(.
Im happy that you still could find the time to test it an update is allready tested as we speak. There is nothing that can be done about the ringing the next version will lower it a little but only with qpel off it would be "almost" invisible, but then you lose details and with the resolution you are useing it would get really blurry, without qpel. Im not sure useing GMC was a good idear it compensates qpels sharpening effect a little so it makes everything blurrier.

[Andrey]

Thank you for kind words... Damn cancer. Really hate it. :devil:

>>but then you lose details and with the resolution you are useing
>>it would get really blurry, without qpel.
Yeah. May be I should stuck with lower or original resolution ?
I've always use ~75% of original when bitrate is considerably low...

>>Im not sure useing GMC was a good idear it compensates qpels
>>sharpening effect a little so it makes everything blurrier.
I did two tests - with and w/o gmc, but one w/o gmc was lost, sorry. From my perspective they were rather similiar, the only thing I have noticed - scene with Obi Van fighting Fett on the station with heavy rain - with gmc is too blurried, w/o gmc is too blocked :)
EDIT:
Forgtot to mention - I did AVC encode too - funny - sometimes it looks better, but some moments are ugly too, and may be even uglier. Jumping blocks - the main problem. Many macroblocks seen. And with postprocessing - blurried to hell...
Sorry, have no tool to cut it from .mp4 and post examples...

[redfordxx]

Hi,

in the beginning you wrote this is optimized for 1.85, 2.35, 2.40.
But what are the resolution for that? Exactly calculated, usually one axis is floating point number... Does it have to be a "multiple of something"?

You are speaking here often about 720x???, but often DVDs are cropped a little, eg i have now LOTR TT EE 712x424 original (or better 420, because 2 lines on top and bottom are quite ugly) I am willing to test it (probably then with SSIM with luma weight). How should I resize it - I am going to cca 1400 kbps video?

I have now XviD 1.1 and you write your tweaked one is 1.0? I don't know much about it, but I suppose, there is something better in 1.1. So, will you follow XviD upgrades to "keep the pace"?

R.

[Sharktooth]

712? i have the TT EE (R2) and it's 720x424... btw i dont resize, i just set the PAR in xvid.
Doing so you get an anamorphic encode (like the DVD) and you dont loose any horizontal line.

[redfordxx]

Originally posted by Sharktooth
712? i have the TT EE (R2) and it's 720x424... btw i dont resize, i just set the PAR in xvid.]
I have 4pix left and 4pix right black. I have the Easter Europe Edition, but i suppose only the audio should be different.
Moreover, I don't believe it is possible to make 720x424@25fps@1400kbps good quality. I tested 712x416crop 1pass, I can't remember now but definitely it would be less than 50%compression.

Doing so you get an anamorphic encode (like the DVD) and you dont loose any horizontal line.
I also plan anamorphic, but resized, because of reasons above. Moreover 720x424 is far from 1.85, so what is the meaning of these numbers in the beginning of this thread?
BTW: For me, PAR in XviD does not work, i use Matroska.

[Sharktooth]

Uhm, if you have 4px left and right it's not the same video.
Btw, i didnt read you would have used 1400kbps...
However if you use the latest ffdshow build ( http://ffdshow.sourceforge.net/tikiwiki/tiki-index.php?page=Getting+ffdshow ), AR will work.
720x424 is the resolution and 1.85:1 is the AR. So if you set the AR you should not care about the resolution coz it will ALWAYS respect the AR you set :)

[redfordxx]

Originally posted by Sharktooth
So if you set the AR you should not care about the resolution coz it will ALWAYS respect the AR you set :)
Well, I understand, what is the difference between AR and resolution.
But, I thought and think, that encoding quality is not related to AR but to resolution..
The when i encode 720x424video, the result should be the same whatever PAR i enter.

Maybe I misunderstood this in the beginning...???

Q: What is it optimized for ?

A: The Extra_Detail Profile (settings) used here were optimized tested and tweaked for
Cinemascope 2.35:1 2.40:1 and 1.85:1 DVD Movie resolutions upto 720x576 D1.

[Sharktooth]

Yes, the higher the resolution the better will be the quality (provided the bitrate is enaugh high).

[redfordxx]

Originally posted by Sharktooth
Yes, the higher the resolution the better will be the quality (provided the bitrate is enaugh high).

Sharktooth,
sorry for bothering and also thanks for patience with me. But I suspect we are not talking about the same world.

When I read the above quoted Q&A, I thought my video resolution should be e.g. 720x300(2.40:1),672x280(2.40:1),592x320(1.85:1) to have benefit of this settings...
So, that was wrong, now I guess. I should just forget that info and use any resolution I like? (multiple of 8????)

[redfordxx]

What are all these Qmatrices you have in the signature about? Is somewhere any description?

[Sharktooth]

Originally posted by redfordxx
Sharktooth,
sorry for bothering and also thanks for patience with me. But I suspect we are not talking about the same world.

When I read the above quoted Q&A, I thought my video resolution should be e.g. 720x300(2.40:1),672x280(2.40:1),592x320(1.85:1) to have benefit of this settings...
So, that was wrong, now I guess. I should just forget that info and use any resolution I like? (multiple of 8????)
Well, if you're going to use PAR in the codec then the decoder will resize the picture to match the AR you specified.
Be carefull with resolutions though. Coz the decoder will resize the horizontal res (ie 720 pixels) to match the monitor res (ie 1024 pixels) and the vertical res (ie 432 pixels) to match the AR (ie 432 pixels).
So the final output will be corrected to 1024x432 if you encoded @ 720x432 but if you encoded with a different horizontal res, the decoder may not work as expected. So try to not resize at all when encoding anamorphically.

Originally posted by redfordxx
What are all these Qmatrices you have in the signature about? Is somewhere any description?
There is a thread speaking of them. Just search for "EQM V3 series" with forum search function.

EDIT: Added the direct link to the thread in my sig.

[CruNcher]

Version 1.4 available
+ changed to Foxers RC
+ added 3 unrestricted Main Profiles (Good,Better,Best)
+ Skals updated Fdct/Idct (SSE2)
+ changed to a unified build (still needs testing)
+ cartoon mode save now

Their 3 unrestricted (non vbv) Profiles now that can be used
Good (Fastest) = The fastest should be used all the time Speed/Quality Balanced (should work on every Standalone Generation)
Better (Fast) = Even more Details are Preserved (useing Qpel) (should work on 2nd Generation Standalones)
Best (Slow) = For extreme compression needs with very shaky movies + very long runtimes (needs a GMC 3 Warp Point capable Standalone)

All your feedback is welcome (source,runtime(min),resolution,profile,chosen bitrate or size,final bitrate or final size,1st pass encoding time,2nd pass encoding time,processor speed in real Mhz/Ghz and core)

Example = 2010DVD,116,704x324,Good,1000,998,00:52h,01:17h,AMD 64 2Ghz Venice

Standalones tested so far

Medion 7475 (MTZ firmware) (MT1389GE) (Good,Better) (Anamorph compliant)
Siemssen SCO 5000 ND (2nd Firmware) (Vadis 776) (Good,Better) (.avi non Anamorph compliant)

[CruNcher]

Please some Feedback it can't be that everyone is happy can it ?

[Sharktooth]

eh... have really no time until the next week :(

[yaz]

yep ... same her ... i would jump onto testing immediately if someone stopped the big-wheel (~time).

thx
y

[Soulhunter]

Im also busy atm... ^^;


Bye

[Sirber]

Sorry, busy :(

[boombastic]

i encoded the last 10 minute of Matrix Revolutions with version 1.4 and good profile.I played it on my DVD recorder which is also a divx player, i don't know the cjip, and there were no problem of compatibility.The quality was also very good!Keep on with theis good work!

[DeathTheSheep]

Woah... I think the no-feedback problem is just a general "Lack of Knowledge" problem: I had no idea that something like this even existed, much less was up for testing... Holy cow. Maybe if you gave it a catchier name like "SuperXviD beta test" or something... :D OK, I'll give it a shot ;)

[SteMa]

Hi, I'm about to make an encode for all matrix movies (1,animatrix,2,3), and get it all to 1 dvd. 2 languages with ogg 5.1 @~192-224kbit (enough?). Than there should be 1100-1300kbit left for the video (dunno how much closing titles). Which resolution I should use? Original, or 640x272? And what settings should I set? Should I use edp 1.4? Would it gain quality?
Or use x264 instead? Does that give more detail/quality? (I can leave my computer on for a weekend in the students hostel). If it does, I would certainly want to use that, but with which settings? :)

[mod]

Hi all.
I've installed the latest (1.4) version to make some comparison between various builds.
I hadn't read any comment about the realtime capture (maybe because this is not what this build was made for :) ) from PAL material.
I've used VDub 1.6.11, and 2 resolutions, a 720*576 and a 320*240, with 800 Kbps as target bitrate (it's low but that's it).
I'm having a big problem: the bitrate remains really low (never reaches 50 Kbps!!)
Any idea? I repeat this was just to make some comparisons.

[CruNcher]

@ SteMa
you have to try that yourself no one can help you with that it's really a very subjective thing (especialy with the usage of inloop deblocking) and still alot tweaked stuff but if you decide for EDP you should take the original resolution, croop away the black boarders and use AR signaling.

@mod
Jep it was never tested under that conditions and could lead to very bad results indeed, another possibility would be you missed to load defaults when testing the differnt builds ?

[CruNcher]

update

Extra Detail Profile Public Test 1.4.5

changelog
+ SSE2 Code sad16/dev16/fdct only used in the Intel build
+ removed Best Profile
+ two builds now AMD and Intel
+ synced with CVS Head 1.2

[BigDid]

update
Extra Detail Profile Public Test 1.4.5...
+ SSE2 Code sad16/dev16/fdct only used in the Intel build
+ removed Best Profile
+ two builds now AMD and Intel
+ synced with CVS Head 1.2
Happy new year Cruncher,

Thanks for the new build, also thanks for cvs synched.

Out of curiosity, are the 2 builds and SSE2...only in Intel build related to the compiler used? if so could you give to a noob like me more infos?

My best wishes

Did

[CruNcher]

Happy new year Cruncher,

Thanks for the new build, also thanks for cvs synched.

Out of curiosity, are the 2 builds and SSE2...only in Intel build related to the compiler used? if so could you give to a noob like me more infos?

My best wishes

Did

Thx also to you a Happy new Year
and to answer your question no it's not only compiler related the SSE2 Code for sad16/dev16/fdct is not active for the AMD build, because it slows down AMD architecture that support SSE2 (AMD64).

[BigDid]

Thanks, seems Athlon64 is more and more the good choice :)

Did

[DeathTheSheep]

I hope I'm not beating a dead horse here.

I'd like to try your CruNcher EDP against the new CVS build, since you advised me to do so back in the DivX thread--this might be just what I'm looking for. Could you sync this with the latest CVS for comparison, since yours is about 2 years old now? Thanks!

[johnsonlam]

Slowly i beginn to fear that nobody is interested in that topic more into other ehm useless things well ok your choice.

Not really, thanks for your work.

I want to test but I'm stuck in work, too busy to really have a test.

[CruNcher]

Hi i see their is still interest in that part of my Research im currently not working on this anymore and the Reason for that is indeed H264/AVC and it's superior Quality (less B-frame problems that i tried to workaround with EDP, target bitrate is more often hit almost 100% no real under or overflow anymore that i tried to compensate with EDP) and most important also tweakability into the old Look and Feel of ASP (closer looking to the original input) but keeping most of it's Compression advances and even faster Encoding @ the same time :) so tell me 1 good Reason why i should continue with ASP if the same Goal can be reached with H264 (X264) and it's better compression Results and especialy wide range Hardware support, with some visual tweaking (that's my current Research) :)

[johnsonlam]

so tell me 1 good Reason why i should continue with ASP if the same Goal can be reached with H264 (X264) and it's better compression Results and especialy wide range Hardware support, with some visual tweaking (that's my current Research) :)

The only reason is someone still want the keep maximum compatibility with the old DivX/XviD hardware players.

The official XviD stopped for a long time since X264 becoming more popular, and now FFDShow can process X264 correctly, so people keep moving on ...

Thanks for your work, nothing wrong if you decided to stop here, maybe you also want to work on something new.

Thanks for your effort!

[DeathTheSheep]

Since you consider this project dead, please provide your sources in .diff format (diff against your CVS or current CVS) so I can look into it.