Results of some Alt.CC testing

[MoonWalker]

I have done some testing for the aggresion of XviD...I have done Matrix..

The settigs where :

XviD-22052002-1 by Koepi
Desired file size : 581625
Motion : 6
Quant : H.263
(After reading this thread http://forum.doom9.org/showthread.php?s=&threadid=26940 I decided to leave the quantizers to defaults).
I-Frame : 2-31
P-Frame : 2-31
High : 300%
Low : 100%
Strengh : 60%
Credits : quant31

Low Aggresion :
DebugView analyzer for XviD codec v0.51 by MoonWalker
e-mail : s_ilias@gmx.net


Quantizers Analisis
---------------------

Quantizers Used For Movie :
------------------------------
Quant 2 Used : 6903 Times, Percentage Used : 3.71%
Quant 3 Used : 87996 Times, Percentage Used : 47.26%
Quant 4 Used : 71427 Times, Percentage Used : 38.36%
Quant 5 Used : 17096 Times, Percentage Used : 9.18%
Quant 6 Used : 2028 Times, Percentage Used : 1.09%
Quant 7 Used : 403 Times, Percentage Used : 0.22%
Quant 8 Used : 51 Times, Percentage Used : 0.03%
Quant 9 Used : 81 Times, Percentage Used : 0.04%
Quant 10 Used : 22 Times, Percentage Used : 0.01%
Quant 11 Used : 61 Times, Percentage Used : 0.03%
Quant 12 Used : 32 Times, Percentage Used : 0.02%
Quant 13 Used : 34 Times, Percentage Used : 0.02%
Quant 14 Used : 13 Times, Percentage Used : 0.01%
Quant 15 Used : 27 Times, Percentage Used : 0.01%
Quant 16 Used : 14 Times, Percentage Used : 0.01%

Average Quantizer Used for Movie : 3.586

Quantizers Used For Credits :
--------------------------------
Quant 31 Used : 9967 Times.


Intra-Frame (Key-Frame) Quantizers
------------------------------------

Movie
-------
Quant 3 Used : 3352 Times, Percentage Used : 98.56%

Credits
---------
Quant 31 Used : 49 Times, Percentage Used : 1.44%

Inter-Frame (P-Frame) Quantizers
------------------------------------

Movie
-------
Quant 2 Used : 6903 Times, Percentage Used : 3.58%
Quant 3 Used : 84644 Times, Percentage Used : 43.91%
Quant 4 Used : 71427 Times, Percentage Used : 37.06%
Quant 5 Used : 17096 Times, Percentage Used : 8.87%
Quant 6 Used : 2028 Times, Percentage Used : 1.05%
Quant 7 Used : 403 Times, Percentage Used : 0.21%
Quant 8 Used : 51 Times, Percentage Used : 0.03%
Quant 9 Used : 81 Times, Percentage Used : 0.04%
Quant 10 Used : 22 Times, Percentage Used : 0.01%
Quant 11 Used : 61 Times, Percentage Used : 0.03%
Quant 12 Used : 32 Times, Percentage Used : 0.02%
Quant 13 Used : 34 Times, Percentage Used : 0.02%
Quant 14 Used : 13 Times, Percentage Used : 0.01%
Quant 15 Used : 27 Times, Percentage Used : 0.01%
Quant 16 Used : 14 Times, Percentage Used : 0.01%

Credits
---------
Quant 31 Used : 9918 Times, Percentage Used : 5.15%



Frame Analisis
----------------

Number Of Intra-Frames (Key-Frames) : 3401
Number Of Inter-Frames (P-Frames) : 192754
Total Number Of Frames : 196155

1.73% of the Movie is Intra-Frames (Key-Frames)
98.27% of the Movie is Inter-Frames (P-Frames)



Size Analysis
----------------

1-Pass Size : 1202443293 Bytes or 1174261 KBytes
Scaled Size : 590220680 Bytes or 576387 KBytes
Actual Size : 587267253 Bytes or 573503 KBytes


Medium Aggresion :


DebugView analyzer for XviD codec v0.51 by MoonWalker
e-mail : s_ilias@gmx.net


Quantizers Analisis
---------------------

Quantizers Used For Movie :
------------------------------
Quant 2 Used : 8292 Times, Percentage Used : 4.45%
Quant 3 Used : 90789 Times, Percentage Used : 48.76%
Quant 4 Used : 64633 Times, Percentage Used : 34.71%
Quant 5 Used : 19015 Times, Percentage Used : 10.21%
Quant 6 Used : 2644 Times, Percentage Used : 1.42%
Quant 7 Used : 450 Times, Percentage Used : 0.24%
Quant 8 Used : 62 Times, Percentage Used : 0.03%
Quant 9 Used : 89 Times, Percentage Used : 0.05%
Quant 10 Used : 21 Times, Percentage Used : 0.01%
Quant 11 Used : 60 Times, Percentage Used : 0.03%
Quant 12 Used : 36 Times, Percentage Used : 0.02%
Quant 13 Used : 39 Times, Percentage Used : 0.02%
Quant 14 Used : 14 Times, Percentage Used : 0.01%
Quant 15 Used : 31 Times, Percentage Used : 0.02%
Quant 16 Used : 13 Times, Percentage Used : 0.01%

Average Quantizer Used for Movie : 3.575

Quantizers Used For Credits :
--------------------------------
Quant 31 Used : 9967 Times.


Intra-Frame (Key-Frame) Quantizers
------------------------------------

Movie
-------
Quant 3 Used : 3344 Times, Percentage Used : 98.56%

Credits
---------
Quant 31 Used : 49 Times, Percentage Used : 1.44%

Inter-Frame (P-Frame) Quantizers
------------------------------------

Movie
-------
Quant 2 Used : 8292 Times, Percentage Used : 4.30%
Quant 3 Used : 87445 Times, Percentage Used : 45.36%
Quant 4 Used : 64633 Times, Percentage Used : 33.53%
Quant 5 Used : 19015 Times, Percentage Used : 9.86%
Quant 6 Used : 2644 Times, Percentage Used : 1.37%
Quant 7 Used : 450 Times, Percentage Used : 0.23%
Quant 8 Used : 62 Times, Percentage Used : 0.03%
Quant 9 Used : 89 Times, Percentage Used : 0.05%
Quant 10 Used : 21 Times, Percentage Used : 0.01%
Quant 11 Used : 60 Times, Percentage Used : 0.03%
Quant 12 Used : 36 Times, Percentage Used : 0.02%
Quant 13 Used : 39 Times, Percentage Used : 0.02%
Quant 14 Used : 14 Times, Percentage Used : 0.01%
Quant 15 Used : 31 Times, Percentage Used : 0.02%
Quant 16 Used : 13 Times, Percentage Used : 0.01%

Credits
---------
Quant 31 Used : 9918 Times, Percentage Used : 5.15%



Frame Analisis
----------------

Number Of Intra-Frames (Key-Frames) : 3393
Number Of Inter-Frames (P-Frames) : 192762
Total Number Of Frames : 196155

1.73% of the Movie is Intra-Frames (Key-Frames)
98.27% of the Movie is Inter-Frames (P-Frames)



Size Analysis
----------------

1-Pass Size : 1202443293 Bytes or 1174261 KBytes
Scaled Size : 590220680 Bytes or 576387 KBytes
Actual Size : 587265438 Bytes or 573501 KBytes

High Aggresion :

DebugView analyzer for XviD codec v0.51 by MoonWalker
e-mail : s_ilias@gmx.net


Quantizers Analisis
---------------------

Quantizers Used For Movie :
------------------------------
Quant 2 Used : 10058 Times, Percentage Used : 5.40%
Quant 3 Used : 92225 Times, Percentage Used : 49.53%
Quant 4 Used : 59141 Times, Percentage Used : 31.76%
Quant 5 Used : 20224 Times, Percentage Used : 10.86%
Quant 6 Used : 3574 Times, Percentage Used : 1.92%
Quant 7 Used : 552 Times, Percentage Used : 0.30%
Quant 8 Used : 79 Times, Percentage Used : 0.04%
Quant 9 Used : 102 Times, Percentage Used : 0.05%
Quant 10 Used : 18 Times, Percentage Used : 0.01%
Quant 11 Used : 66 Times, Percentage Used : 0.04%
Quant 12 Used : 36 Times, Percentage Used : 0.02%
Quant 13 Used : 44 Times, Percentage Used : 0.02%
Quant 14 Used : 16 Times, Percentage Used : 0.01%
Quant 15 Used : 37 Times, Percentage Used : 0.02%
Quant 16 Used : 16 Times, Percentage Used : 0.01%

Average Quantizer Used for Movie : 3.568

Quantizers Used For Credits :
--------------------------------
Quant 31 Used : 9967 Times.


Intra-Frame (Key-Frame) Quantizers
------------------------------------

Movie
-------
Quant 3 Used : 3328 Times, Percentage Used : 98.55%

Credits
---------
Quant 31 Used : 49 Times, Percentage Used : 1.45%

Inter-Frame (P-Frame) Quantizers
------------------------------------

Movie
-------
Quant 2 Used : 10058 Times, Percentage Used : 5.22%
Quant 3 Used : 88897 Times, Percentage Used : 46.11%
Quant 4 Used : 59141 Times, Percentage Used : 30.68%
Quant 5 Used : 20224 Times, Percentage Used : 10.49%
Quant 6 Used : 3574 Times, Percentage Used : 1.85%
Quant 7 Used : 552 Times, Percentage Used : 0.29%
Quant 8 Used : 79 Times, Percentage Used : 0.04%
Quant 9 Used : 102 Times, Percentage Used : 0.05%
Quant 10 Used : 18 Times, Percentage Used : 0.01%
Quant 11 Used : 66 Times, Percentage Used : 0.03%
Quant 12 Used : 36 Times, Percentage Used : 0.02%
Quant 13 Used : 44 Times, Percentage Used : 0.02%
Quant 14 Used : 16 Times, Percentage Used : 0.01%
Quant 15 Used : 37 Times, Percentage Used : 0.02%
Quant 16 Used : 16 Times, Percentage Used : 0.01%

Credits
---------
Quant 31 Used : 9918 Times, Percentage Used : 5.14%



Frame Analisis
----------------

Number Of Intra-Frames (Key-Frames) : 3377
Number Of Inter-Frames (P-Frames) : 192778
Total Number Of Frames : 196155

1.72% of the Movie is Intra-Frames (Key-Frames)
98.28% of the Movie is Inter-Frames (P-Frames)



Size Analysis
----------------

1-Pass Size : 1202443293 Bytes or 1174261 KBytes
Scaled Size : 590220680 Bytes or 576387 KBytes
Actual Size : 587330004 Bytes or 573564 KBytes

Well, as I can see the average quantizers doesn't floats too much..If we want some blind results at these settings, medium has better quant the low and finaly high..The other thing is that with medium and high we have reached quant16 whereas with low we have quant14...


MoonWalker

[Jon Ingram]

You've probably spent quite a lot of time on this, but you're testing the wrong thing. It doesn't matter what the distribution of quantisers is -- what matters is how the encoded movie looks when you watch it.

Why don't you look at the clips you've made, and give us some subjective judgements on the quality? Even if you can only distinguish a few as being obviously bad, that would be helpful.

[MoonWalker]

Sorry for posting so late my visual results...

Well without any post-proccesing..

At the (famous) lobby scene with low there aren't so much blocks when wacthing the movie, medium has more block and high has ever more block(especialy when there is smoke) and I noticed some wierd moving of the wall when they are covered with smoke..Like shifting a bit.

The scene when Neo says :"I want guns,lots of guns" and the guns are running to the backround is typical to see what the aggression does..With Low there are less block but they last a little bit longer.With medium there are virtualy the same blocks like low but last a not so much..And high was the blockiest but the duration was very small..

Now with post-proccessing(with Nic's DSF all enabled) with low I hardly could see any block and with High there where very little blocks...

Just to say that this test was on video and not in pictures.If I had looked at the frames, well it would be much much worst..Besides we are dealing with a video compression codec :)...

If you want me to do some other tests just say it..

MoonWalker

[Teegedeck]

Thanks for posting your impressions! They resemble my own experiences. :)

[OUTPinged_]

MoonWalker, it was a bad idea to not cap quantizers with your setup.

2-6 range for your encodes would been producing better quality in critical parts of the movie.

[MoonWalker]

Originally posted by OUTPinged_
MoonWalker, it was a bad idea to not cap quantizers with your setup.

2-6 range for your encodes would been producing better quality in critical parts of the movie.

You mean I-frame quants only or both??
Personaly I always do 2cd encoding(cause I keep the Ac3 track) and my settings are I-frame : 2-3 P-frame : 2-4(or 2-5 for a 2:10+ movie)

MoonWalker

[Acaila]

MoonWalker, it was a bad idea to not cap quantizers with your setup. 2-6 range for your encodes would been producing better quality in critical parts of the movie.
I thought we had decided to use the full range for testing purposes?? :)

[MoonWalker]

I thought we had decided to use the full range for testing purposes?? :)

Well that's why I leave XviD decide the quantizers..I usualy cap quantizers but I read this http://forum.doom9.org/showthread.php?s=&threadid=26940 and decided to leave tha codec to decide..


MoonWalker

[Teegedeck]

I think we weren't finished, yet...:)

So, anyone who found medium aggression gave him a nicer picture than low aggression in a direct comparison (same movie)? If not, I'd say we agree that low aggression is a recommendable standard setting for now.

OK?

[NoFace319]

humm... I find that encoding with a low agression makes the Panoramic scenes look bad a little bit; specially when the scene is little dark... for xample I tryed 2 encode American Beauty... a hard one... don't know but I've been testing with all of parameters and allways Blocky (not 2 much but... bothers) and finally I have decided that if I leave At. Curv "OFF" the film is much more compesated and has a constant quality... Don't know but listen I'm eagger to find some standart setting for Att. Curv. cos If they R there is cos they can b useful :)

Actually I never use Att. Curv.

Anyway... Anybody has encoded Saving Private Ryan??? I think that THERE the Att. Curv. MUST BE USED... I'm right? So can you help me on this?? :)

U know... I think that we must creat a topic to post our Settings and Parameters when we encode our movies cos I see that most of our time is spent in TESTING!!! ok... also is learning but... sometimes we want encode and go to bed... do u agree with me?? So.. I'll make my firs step right now.

[Rhaegar Targaryen]

I've done a few movies so far, and all looked great at Low agression but I finally came to one which looks much better at Medium; True Romance. FYI.

[iago]

@Teegedeck

I think we weren't finished, yet... And imho, I guess we still AREN'T finished yet! :)


New builds are being released day by day (Koepi is really fast nowadays :)), and now we (happily) have a few more parameters to deal with. All right. But what about our recommendable, generally-agreed-upon standard settings discussion regarding altCC? Or has everybody already hardcoded (or is totally certain about) their own altCC settings? :) Well, I still am not, except the one Teegedeck points out.

"...I'd say we agree that low aggression is a recommendable standard setting for now. OK?" So far, at least for me, OK! :)

Concerning the high/low distance settings, can high300/low100 be a compromise for "most" movies?

Or, high150/low50 for "low bitrate" and "dark" ones?

Although recently most people seem rather unwilling to discuss these any more, I'd be very glad if we continue to share our opinions and experience on these parameters.

Best regards,
iago

[Koepi]

I'm going to set some new default-settings which work nicely with almost all movies.

I'm still not convinced that all these tweaking-possibilities really help the encodings, you can do MUCH harm with these settings.

But for now I'm testing:

Medium curve aggression
Payback delay: 250
High distance: 500
Low distance: 90

Automatic min qual
strength: 30%

The results look promising, but still aren't what I expect from curve treatment.

Ah, and set iframe boost to 0.

Regards,
Koepi

[iago]

@Koepi

I couldn't help making a test encode using the 08082002 build (Fight Club again, 640*272, neutral bicubic, and 965900 kb aimed video size for 2CDs with AC3 audio - to achieve the best possible encode of this movie), with the altCC settings you mentioned :). I must admit that the result is the best one I've got so far with this movie, compared to my previous 2CD (with AC3 audio) rips both with the older builds of xvid and other codecs (sbc, divx502Pro) as well. :)

Also, I'm no more sceptical about using lumi masking in both passes, because in my recent encodes I've seen that using lumi masking in both passes really works like a magic. :)

Well, for the time being, I'm very pleased with the results I get with the 08082002 build, and even more after trying your experimental altCC settings.

Still, I have one question in mind: Why do you set I-frame boost to 0? Do you think a higher boost (eg. a boost of 20%) steals too many bits from P-frames, because of the possibility of getting too many consecutive I-frames in the movie (I remember discussing this issue just in the opposite way formerly - a quality boost in the whole scene with higher I-frame boost)?

Anyway... Thanks a lot for your ongoing effort, your support, and for this very very promising build! :)

Best regards,
iago

[NSanity]

I just did my first Alt.CC encode then (Chapter 29 of LOTR:FOTR R4) using


AVS file
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
mpeg2source("C:\Program Files\DVD2SVCD\Movie\DVD2AVI_PROJECT_FILE.d2v")
crop(0,75,718,424)
Sharpen(0.40)
BicubicResize(640,260,0,0.70)

Xvid 08082002-1 by Koepi
Desired file size : 89000 (roughly 1500kbps)
Motion : 6
Quant : Modulated
I-Frame : 2-31
P-Frame : 2-31
High : 300%
Low : 100%
Strengh : 60%
luma-masking both passes
Medium Alt.CC

looked fairly grainy without post-processing (i couldn't turn it on if i wanted too... )

Trying the following

AVS file
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
mpeg2source("C:\Program Files\DVD2SVCD\Movie\DVD2AVI_PROJECT_FILE.d2v")
crop(0,75,718,424)
BicubicResize(640,260,0,0.70)

Xvid 08082002-1 by Koepi
Desired file size : 89000 (roughly 1500kbps)
Motion : 6
Quant : H.263
I-Frame : 2-31
P-Frame : 2-31

Medium curve aggression
Payback delay: 250
High distance: 500
Low distance: 90
Automatic min qual
strength: 30%

will report back in 30 mins...

if it looks good, off we go with a 3cd rip of lotr:fotr :)

[NSanity]

well the grainy-ness is gone (sharpen i spose), and the colours are a fair bit sharper (is this what luma-masking does?), whilst the stills of people with motion in the background are fairly sharp, how can i get them (the focus) slightly sharper? suggestions?

[iago]

Hi everybody,

I'm planning to do an interesting :) 1CD test encode with Koepi's 08082002-1 build, after which I'll compare the results with the 2CD (with AC3 audio) Nandub SBC rip of the same movie, especially in terms of blockiness in dark areas.

Well, I know it is not fair indeed :), but the reason for doing such an unjust comparison is that I actually hope to see the XviD codec do its magic against my old love nandub SBC, even under unfair, unequal conditions :). Anyway, whatever the result will be, possibly it will be a no win - no lose comparison in the end, since the circumstances are not equal! (A comparison under equal conditions!? I guess the winner is already clear... :))

So, the movie is The Score (1hr 59min), a dark and well compressible one. Settings are 640*272, neutral bicubic, in both cases. The SBC (2CD) encode is 1032000 kb video, and XviD (1CD) encode will be 640000 kb video.


Xvid
First pass:
6.UltraHigh
H.236
Lumi masking enabled

Second pass:
6.UltraHigh
Modulated
Lumi masking enabled
I-frame boost: 0 (koepi said so :))
(Max-Min I-frame intervals: 250-1)
Min-Max I-frame: 2-4
Min-Max P-frame: 2-8
AltCC low/high250/low50
all the rest default

Any suggestions?
Well, it will be interesting, won't it?
And... am I expecting too much from XviD? ;)

best wishes,
iago

[Koepi]

a) you're expecting too much (well, the quality should still be in range. but magic isn't working all the time...)
b) use the curve settings i provided above. You will most certainly destroy your encoding with other settings.

[iago]

@Koepi

Yes, you are absolutely right, I know it will be too harsh a comparison under such unequal conditions. Just some (maybe pointless :)) try out of sheer curiosity. But why not give it a go?

And all right again, from now on:

Medium curve aggression
Payback delay: 250
High distance: 500
Low distance: 90
Automatic min qual strength: 30%
I-frame boost: 0

:)

best regards,
iago

[iago]

Hi Koepi and everybody,

Here are the results of the unjust :) test I mentioned above.
The comparison of the Nandub-sbc (2CD with AC3 audio) and Xvid (1CD with ogg) encode of the same movie.

SBC: 1032000 kb video (average quantizer ~ 2.5)
Xvid: 640000 kb video (average quantizer = 3.389)

Movie: The Score (119 min)
640*272 / neutral bicubic

Xvid

Xvid
First pass:
6.UltraHigh
H.236
Lumi masking enabled

Second pass:
6.UltraHigh
Modulated
Lumi masking enabled
I-frame boost: 0
Max-Min I-frame intervals: 250-1
Min-Max I-frame: 2-4
Min-Max P-frame: 2-8
AltCC medium/high500/low90
strength: 30
payback delay: 250


------------------------------------------------------
Results:

DebugView analyzer for XviD codec v0.10 by MoonWalker
e-mail : s_ilias@gmx.net

Quantizers Analisis
---------------------

Quantizers Used For Movie :
------------------------------
Quant 2 Used : 2070 Times, Percentage Used : 1.22%
Quant 3 Used : 101461 Times, Percentage Used : 59.58%
Quant 4 Used : 65240 Times, Percentage Used : 38.31%
Quant 5 Used : 1408 Times, Percentage Used : 0.83%
Quant 6 Used : 101 Times, Percentage Used : 0.06%
Quant 7 Used : 2 Times, Percentage Used : 0.00%
Quant 8 Used : 1 Times, Percentage Used : 0.00%

Average Quantizer Used for Movie : 3.389

Quantizers Used For Credits :
--------------------------------
Quant 31 Used : 8293 Times.

MPEG Quantization Type Used 111824 timed, Percentage Used : 62.62%

H.263 Quantization Type Used 66752 timed, Percentage Used : 37.38%

Quantizers prevented from rising too steeply 0 times


Intra-Frame (Key-Frame) Quantizers
------------------------------------

Movie
-------
Quant 3 Used : 1207 Times, Percentage Used : 87.02%
Quant 4 Used : 138 Times, Percentage Used : 9.95%

Credits
---------
Quant 31 Used : 42 Times, Percentage Used : 3.03%

Number Of Consecutive I-Frames : 50

Inter-Frame (P-Frame) Quantizers
------------------------------------

Movie
-------
Quant 2 Used : 2070 Times, Percentage Used : 1.17%
Quant 3 Used : 100254 Times, Percentage Used : 56.58%
Quant 4 Used : 65102 Times, Percentage Used : 36.74%
Quant 5 Used : 1408 Times, Percentage Used : 0.79%
Quant 6 Used : 101 Times, Percentage Used : 0.06%
Quant 7 Used : 2 Times, Percentage Used : 0.00%
Quant 8 Used : 1 Times, Percentage Used : 0.00%

Credits
---------
Quant 31 Used : 8251 Times, Percentage Used : 4.66%



Frame Analisis
----------------

Number Of Intra-Frames (Key-Frames) : 1387
Number Of Inter-Frames (P-Frames) : 177189
Total Number Of Frames : 178576

0.78% of the Movie is Intra-Frames (Key-Frames)
99.22% of the Movie is Inter-Frames (P-Frames)



Size Analysis
----------------

1-Pass Size : 1210649494 Bytes or 1182274 KBytes
Scaled Size : 651058108 Bytes or 635798 KBytes
Actual Size : 651055496 Bytes or 635796 KBytes

Usefull Statistics
------------------
Compressibility : 53.78%
Relative Quality of XviD avi : 59.01%
Absolute Quality of XviD avi : 95.83%

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

And finally, as I stated above, the major (but not only) aim of this test was to compare the two encodes of the same movie (with dark scenes abundant) in terms of blocks. And in my opinion XviD 08082002 (with the AltCC parameters used) was very successful in the test. The visual quality (at least for my eyes) is no less (but generally even better) than the 2cd sbc encode throughout the whole movie. :)

Also, I'm providing below some randomly taken screenshots (in VirtualDub with no post-processing) for those who are interested.

best regards,
iago

[iago]

Sorry, I forgot to say that the screenshots include both sbc and xvid encoded frames to provide a comparison between the two.

iago

[Emp3r0r]

wow, great test, XviD did a good job judging by the screenshots and the unfair test

[Sigmatador]

wow! xvid 0.167 vs sbc 0.270 bit/frames*pixel, the result --> xvid looks great and without postprocessing ^^ i think i prefer xvid screenshots :cool:
doom9 have to update the comparaison :D
some questions plz:
is your parameters good for all movies? is medium curve agression better than high or low? and what about the matrix and ryan screenshots of the comparaison with this parameters? (i hope xvid will kick rv9 and divx502)

[iago]

@Sigmatador

You know Doom9 added some updated info concerning XviD's bad performance in his Matrix and SPR test. I guess in his next test with a "new trusted build" XviD will display its real performance.

As for the alt.CC parameters, these are what Koepi suggested to use.

I haven't made tests with Matrix and SPR ('cause I haven't got them), but I'll try to do other tests with some different movies, after which I'll send screenshots again.

so long,
iago

[OUTPinged_]

@Koepi:


On the medium vs low aggression settings debate:

My personal experience was that low aggression can be dodgy on some movies that pocess great framesize range in case you are setting great high distance ( 500, hehe).

low agg:

Pluses:
1. quantizers deviate pretty weakly and are producing steady mediocre quality encode.
2. extremely large and ugly frames are compressed pretty well if high distance is large enough.

Minuses:
1. You have to be very carefull with quantizers going too high on large (300+%?) framesizes. Probably capping max. quant. will help. Good understanding of min_rel_qual will help too =)

2. It just doesnt work good when all frames are about the same size. Roughly half of a movies i encode have pretty small fraemsize deviation and 90% of frames are in 200% high distance range. 500% with low aggression will result in boring encode with quantizer deviating by 2-3 at most.

medium agg:

Pluses:
1. Extremely predictable. Acts familiar to all regular CC users. Newbie friendly.

2. Doesnt care about distance much, works okay with all values. High values should be compensated by right mrq.

Minuses:

1. "Extremely large frame" issue cant be solved by cc in that case.

2. Linear framesize relation to quantizer is just not right. :-( If motion masking will ever be implemented, low distance will surely be a winner. As for now, medium agg. produces _worse_ results in general. But that is only true when you have both mrq and max. quant. set right. People dont care about resulting mrq much and that is bad.

[iago]

@Koepi

I have one question please. Before in this thread, did you suggest I-frame boost: 0 for the specific "altCC" parameters you provided (medium aggression/high500/low90/strength30) or do you suggest I-frame boost: 0 for "default two pass" without altCC too?

Thanks in advance,
iago

[Koepi]

my settings apply to altCC, iago.

I can't recommand "old" CC anymore as ther eis no algorithm to take care of extreme situations, whih _can_ result in (plenty) over-quantized scenes.

Just use the settings I provided, and be sure, you can't improve that result more than 0.5%.

Regards,
Koepi

[iago]

@Koepi

Thanks a lot for the immediate reply :).

best regards,
iago

[ookzDVD]

@iago,

I'm still using the default for alt.cc :
medium : high/low : 300/300

Not bad, working for most types of movies,
at least for me ;)