XviD-27072002-1

andy

Hi folks,

Foxer and I had some hours chat again and here's the result, an improved 2pass curve treatment (again) :

Changelog:

XviD-27072002-1:
- EPSZ(^2) Motion Estimation activated
- New curve treatment revised, should work even better now.

You'll find two new fields in 2pass tab.

One where you can enter a threshold below which keyframe occurences will be treated as consecutive keyframes, Those keyframes get some bitrate reduction except the last keyframe of such a series, which gets quot;full qualityquot; again. This should help low bitrate conditions a lot.

Please, test this build, I even took the time to correct the behaviour if you want to naviagte with the tab-button on your keyboard

Best regards,
Koepi

Koepi,
You Are Great !!!   I Finished A Test Last Night , Comparing SBC 2pass , Div5.02 2pass , An Xvid 7-23-02 2pass /with new curve treatment . I used Decomb3.8 , SSHiQ AviSynth Filter , An BicubicResize 512x384 , In An AVS Then Converted Source To HuffYuv . An Then Started Testing . Div5.02 Had Quater Pixel , Bi-Directional Encoding An I Changed Max Quant To 8 . NanDub Used AntiFreeze , 25% moition Based Curve , LumA Correction , Bit Reservoir Modulation Enabled .  Xvid Used Luma Masking ,  H.263 , Hinted ME , An High Alt. Curve Aggresion .  There Was Absoluly NO Comparison , The Newest Xvid Build Is Absolutely The BEST . An I Would Also Like To Thank You , Koepi , For Your Efforts Put Forth An Given To The Testing Community . PS : I Like Your Site Changes !

Wohoo,

I'm impressed

Thank you very much for testing (thank god it went well  ) with these immense efforts, Surreptitous!

I'm glad you like it (as well as my new site layout  ). I think you'll be even more impressed by the new build, but don't feel like you have to test it this way again!

Thanks a million again,

best regards,
Koepi

I Still Have The 16Gig HuffYuv Source Of The Movie `The One` Which I Like To Test With , Cause It Has Alot Of Action , Stop Motion 360 Scenes , Like The Matrix . An Is Only A 1hr30min Long So It Can Easily Fit On 1Disc .  An You Can Bet Ill Test The New Build Out Tonight An See If I Can Tell The Difference Between build 7-23-02 An 7-27-02 .  Keep Up The Good Work .

Koepi: i haven't encoded anything yet with the new build as I already have one going, but I did notice some GUI bugs which are nothing more than visual.  The inset 3D frames are displayed on top of the text.  Previous build was ok.  Just letting you know even though it has no importance whatsoever.

Koepi: i haven't encoded anything yet with the new build as I already have one going, but I did notice some GUI bugs which are nothing more than visual. The inset 3D frames are displayed on top of the text. Previous build was ok. Just letting you know even though it has no importance whatsoever.

yes, i can confirm this.

@ koepi:
thanks for sacrificing friday night in favour of a new,improved build  

i'll test it tomorrow.

thanks for sacrificing friday night in favour of a new,improved build

well, this is a sad story as my GF is on vacation for gt; 2 weeks now and it will take another 6 weeks until she returns. So nothing really left except for work, work and work a little more to confuse my thoughts.

But this is no post that's meant to make you feel sorry, because I have some success story to tell...

I tried to encode R2 Rush Hour 2 at 704x304, heading for 1 CD (using trbarry's simpleresize filter and english burnt in quot;subtitlesquot;, but just the translations of the parts where they're speaking chinese, and using 2 languages as audio, english and german. so the movie shouldn't exceed ~632MB)... and the result is amazing. Nearly no blocks, even in the fast motion scenes, real fancy details (the leather jacket of chris tucker when he goes to the casino and in many places more),...

I think _now_ XviD can really compete very well in the highest league! Even without bframes (they don't give good results either, just take a look at DivX5, the bframes are butt-ugly), qpel (will come soon and improve quality just another bit) and GMC (once more doesn't help quality noticably).

Thanks for your feedback people, let it all out here

Best regards,
Koepi





Everything fine but I can't open Koepi's site. What's going on?

Well done Koepi (love the new site BTW)  Im off for a week now....

(im looking forward to having QPel...  (although it might be a little way off))

-Nic

TAB works!!

Thank you ever so much for taking the time to fix it, Keopi.

Now, I've managed to fit Terminator 2 ESE (544x224) on a single 800MB CD with the July 20th build (I think, anyway: the older build that had consecutive KF handling). No blocks to be seen anywhere. Very acceptable quality. I was very surprised with that since GKnot told me I'd need 2 CDs for it. The first time I get away with disregarding GKnot's recommendations.

Of course, I cannot be sure it's because of the new consecutive KF handling code since DebugView won't run on my PC, for some reason (just hangs if I run it. I already had the latest ver and reinstalling didn't help (no corrupt .EXE, so)), but I plan on some test encodes using older builds once I have some free time.

Will let you know.

Hi Koepi,
Using Mozilla 1.1 beta, I cannot access your web site: I get an quot;Error 404 - File not foundquot; error. I did not have this problem with previous version of Mozilla (1.1 alpha) nor with IE.
BTW, I love your new builds ! 23072002 build did a wonderful job on Ali and I 'm gonna test your new build. Do you intend to add Nic's filter in your package ? (it was not present in 23072002-1 one)
Cheers,

Pasqui.

good work !

is it right, the color of your webpage ist grey / white ?? (opera)
looks strange

hi Pasqui,

I think this is a problem of mozilla (quot;since the day they've put the dl manager control in the settings there have been ZERO changes in mozillaquot; - : showthread.ph...958#post157958)

and as you can read on Koepis site:

This referer-check has some disadvantages: some browsers don't send a referer and you get redirected to an error-page. Plenty people use download managers and didn't switch on to transmit a referer, in such a case the file you want to download doesn't get delivered, too. In such cases try to use another browser or work without a download manager, the files are very small and can be downloaded without problems even for modem users.

guess what?...

I've made some tests this morning and must say that the quantizer-distribution seems to be much-much 'smoother'. Using Matrix and comparing this build aginst the '20072002' shows a clear difference when stepping frame by frame, so this is definately a step in the right direction.

Well done.

@Koepi

I can only say quot;GREAT WORK DONE!quot;. I did a test and converted quot;The Last Castlequot; 124min 2CD / AC3 and there are no Macroblocks! The Quality is really amazing... Perhaps I will try and squeeze it to one CD with MP3...Thx Dude...

CU
EASY123

Some question

1)The min I-frame intreval should be 1 or it doesn't matter..
2)what are the lt;lt;suggestedgt;gt; values for Below i-frame distance and I-frame bitrate reduction? I am thinking 5/20??

Thanks,
MoonWalker

BTW I confirm the bug in the GUI..

Yeah, I have that bug in the GUI as well. That happened as I resorted the tab-order... dunno how to solve it :-(suggestions:
1) Min iframe interval shouldn't matter anymore with this curve treatment, you can set it to one (but you should raise the reduction value then!)

2) I tested the code with iframe distance 10 and reduction 50.

This works very well for one cd encodes.

You have to realize that the code is a little sophisticated: if there _are_ p-frames within consecutive keyframes (well, those we treat like them due to the iframe threshold span above), the reduction is less. if we have the sequence I-P-I , redutcion 50%, iframe distance threshold  10, the first keyframe would still get 75% of the toal bitrate as the p-frame can compensate as well!

I hope this helps,

best regards,
Koepi

Just for clarification...

If we set the threshold to 10, I-frames the are 10 or LESS in distances whould be treated as consecutive and the first one will get the reduction???

MoonWalker

Now it starts getting nerve-wasting...

I-1-2-3-4-5-6-7-8-9-I

That's 10. ok?

Regards,
Koepi