Blocks and more blocks

[SleepingLily]

Hello,

I've tried copying Excalibur twice and both times ended up with a lot of blocks. The more I research this topic, the more the confused I'm becoming. Following the xvid guide... The first attempt was with 1cd and I thought that this maybe the problem since the movie is 2:20 long; but it happened with 2cd's also. I'm thinking that maybe the lumi masking maybe the issue but I'm not certain because it worked so well with Impostor and Dungeons & Dragons. With it, I tried it only in the second pass and then the next attempt I tried it in both passes - same results.

Here are my settings:
(2cd) xvid - 2pass 2nd pass int:
Motion - 6-Ultra High
Quant - H.263
4cc - Xvid
Max I-Frame int - 300
Lumi enabled

2 Pass tab:
I frame - 20%
Discard first pass
Below I frame - 10
I Frame - 30
Rest default values

Alt Curve
Curve Agress. - Med
High Avg - 500 also tried 250 the first time
Low Avg - 90
Strength - 30
Rest default

Quant
Min I-Frame - 2
Max I-Frame - 31
Min P-Frame - 2
Max P-frame - 31

Not sure if this matters but I wasn't able to get the Aspect error to 0.0 and still be within the suggested guidelines of the Bits/Pixel*Frame. (0.216) I went with Bits set at 0.276 with an Aspect error of -0.1, Bitrate 160. Used Sharp Bicubic suggested by GKnot and went with the rest of the values according to the guide.

Thanks for any help,
SleepingLily

[sam_b]

Can't see anything obviously wrong (except I think sharp bicubic is a usually a bad idea for 1cd, but I don't know the film) but I would suggest not using lumi, especially for the 2cd. I, and many people here, prefer linear scaling to altCC: this is likely to cause blocking on high-motion scenes. Can you load your stats file into koepi's StatsReader and find out the first pass size?

Oh, and sorry if I missed something obvious. Never used GKnot.

[Acaila]

I went with Bits set at 0.276 with an Aspect error of -0.1, Bitrate 160.I do hope you made a typo here since a bitrate of 160 is about 1/5 of what most people consider the lowest usable bitrate.

Also, what resolution did you use?

You also might want to start thinking about using smoother filters. GKnot's build-in smoothers work quite well; most people also use Convolution3D nowadays to decrease blocking.

[vlad59]

This is the last rip I did ....

I got a good result using Convolution3d (0, 3, 4, 4, 4, 3, 0) before a soft bicubic (I prefer softness to blocks ;)), Xvid (I don't remember exactly the settings but I'm sure I was capping my quantizers to 2-5 for I frames and 2-10 for P frames) and Vorbis.

I'm sure I haven't used lumi masking at all because it was a bit buggy in august.
My resolution was 640x??? IIRC.

Hope this helps

EDIT : Forgot to say it was a 2 CD Rip with French and English Vorbis audio track

[Koepi]

RefDivX' lumimasking code should be safe to use, so at least with my builds you could definatly give it a try ;)

Best regards,
Koepi

...who is preparing a new build for uploading later tonight...

[NeVeRLiFt]

Originally posted by Koepi
RefDivX' lumimasking code should be safe to use, so at least with my builds you could definatly give it a try ;)

Best regards,
Koepi

...who is preparing a new build for uploading later tonight...

weeeeeeee!

I been having great sucess with your builds and I like what you include with your build.

[iago]

Originally posted by Koepi
...who is preparing a new build for uploading later tonight... Great! I'll give it a real torture test with a hard-to-compress NTSC non-anamorphic hybrid source, using b-frames 4-100-200 and credits with quant 31! ;)

iago

[NeVeRLiFt]

Originally posted by iago
Great! I'll give it a real torture test with a hard-to-compress NTSC non-anamorphic hybrid source, using b-frames 4-100-200 and credits with quant 31! ;)

iago

I think I will use U-571 for my testing :D

[iago]

@NeVeRLiFt,

Maybe I shouldn't ask that but how come you happened to be so interested in and enthusiastic about XviD matters?

The last time I remember you were indulged in some weird fantasies of yours in another thread?! :D

iago
(sorry to shift O/T btw)

[NeVeRLiFt]

I have my reasons. The big one is because Nandub/DivX3.11a needs to be replaced ;)


And to take this even more O/T have you all heard about/tried using iuVCR and the Cyberlink deinterlace filter and videocapturing with XviD?
It has to be seen to be believed... some very sweet quality ;)

[iago]

I have my reasons. The big one is because Nandub/DivX3.11a needs to be replaced ;)

Hehe, I agree; and that's what many people had anticipated long ago you know! :D

Also, I must admit that I'm totally ignorant about the other things you have mentioned in your post! ;)

iago

edit: anyway, back to topic, I'm looking forward to Koepi's new build for the torture test! ;)

[Selur]

ehmm,... Koepi put up a new build,...

see:
http://forum.doom9.org/showthread.php?s=&postid=234836#post234836
At least his post there was after the post in this thread,...

Cu Selur

[iago]

Oh, I see, thanks for pointing this out! ;)

So I have to wait for another new build I guess! ;)

regards,
iago

[Mike456]

What was the source? You can put and 2:20 long movie of an DVD on 1cd and you should have still have good quality (if it's not a high action movie or excremly noisy). And if the source was an old noisy vhs, an super/high 8 video or something like that and you don't use any flters the result is probably blocky even if you put only 1 hour on 1cd

[SleepingLily]

Hello All,

Sam_b: You mentioned linear scaling to AltCC. Where can I find more information about this? I loaded the stats file but I don’t know what I can (or should) do with this information.

Frames – 202453, av.Bits – 0.483, V. Size – 2202226KB
Frame
Mx – 100815
Av – 11138.784
Sd – 8385.824

Motion
mx – 229
av – 46.807
sd – 79.240

Luma
Mx - 0.008
Av – 0.008
Sd – 0.0

First Pass Stats – 57% of 0.483 same with Compress test

Acaila: Bitrate of 160 for 2cd’s was not a typo. I'm a newbie and I went with what the xvid guide suggested. What is an acceptable range without affecting the quality of the video? Is there a formula or guideline to this? (resolution – 576x320)

Vlad59: I downloaded your filter (v1.01) and will read up on it. Thank you for sharing some specs... :)

Mike456: The source is from a dvd. It does have some high action in places and I noticed that even though I selected 2 cd’s, the video is only 689MB. "???"

SleepingLily

[iago]

Acaila: Bitrate of 160 for 2cd’s was not a typo.

@Acaila,

Fortunately, it is the "Audio" bitrate! :D


You mentioned linear scaling to AltCC. Where can I find more information about this?

@SleepingLily

You might want to give my nasty "XviD Two Pass Encoding - Internal Linear Scaling" document a try. It must be somewhere around in this page! :D


regards,
iago

[sam_b]

Iago's doc is essential reading so please read it, but to do a quick test just set curve compression to 0,0,250 and disable altCC. If I can decipher your stats right, you are looking at a definite 2cd output. This is using 2-pass internal mode on the *original* 1st pass stats file, not using any of Gknots scaling. Do this for 10 minutes of the film, abort, and see if it looks better. If it does then you're sorted. If it doesn't then let me know.

But I don't think sharp bicubic is good in this situation, but it's a preference thing.

I apologise again if this makes no sense. Never touched GKnot.

And go read iago's document. It's pretty good. And self-explanatory.

[SleepingLily]

I downloaded your Xvid Two Pass Encoding, iago and read Koepi’s Xvid Options Explained. I started over and some new problems popped up when I tried to load the .avs file. After adding mpeg2dec3.dll and modifying the avs file - LoadPlugin("D:\GORDIA~1\mpeg2dec3.dll",lumoff=-2) upon opening this happened...

Avisynth open failure
Script error LoadPlugin doesn't have a name argument "lumoff"

I removed the lumoff and then received this one -

Unrecognized exception (line 40) which is the resizing
BicubicResize(576,320,0,0.5)

(Virtual Dub 1.4.13) I didn’t do anything different than before, I still have an aspect error of -0.1 and bits/pixels*frame is still 0.276 – 2cds. Any words of wisdom to share?

BTW... Thank you iago for your guide.

SleepingLily

[bilu]

Usage of MPEG2DEC3:

loadplugin("c:\whatever\mpeg2dec3.dll")

mpeg2source("c:\movie\movie.d2v",lumoff=-2)


It's on the Readme file of mpeg2dec3

[sam_b]

Isn't it lumafilter(-2,1) these days, separately?

[manono]

Hi-

Isn't it lumafilter(-2,1) these days, separately?

Yes, if you're using AviSynth 2.5. But SleepingLily's using GKnot and the included AviSynth 2.07, so bilu had it right.

[SleepingLily]

The mistake was the version I was using. The mpeg2dec3.dll I downloaded was for Avisynth 2.5 and when I switch it to the beta 6, everything loaded fine. I forgot to mention that I used neutral instead of sharp this time. I’ll run it tonight.

Also forgot... Koepi if you’re reading this, thank you for your outstanding guide. (Xvid Options Explained) Definitely a must have, very helpful for us newbies. :)

SleepingLily