Is my .AVS ok for LOTR on 2 CDs?

[haazheel]

Hello,

I'm a french new member and for my first post, I want to ask your opinion about that .avs.

I want to encode in DivX5.02 Pro the Lord of the rings, with a ogg sound file.

What do you think of that:

LoadPlugin("D:\Tools\GORDIA~1\mpeg2dec.dll")
LoadPlugin("D:\Tools\GORDIA~1\convolution3d.dll")

mpeg2source("D:\films\DVD\lotr.d2v")

crop(0,65,720,446)

convolution3d(preset="movieHQ")

lanczosresize(640,272)


Do I really need the convolution3d filter?
I have already read that it's more useful for a 1 CD rip...

And for a 2 cd rip, are B-Frame, psychovisual enhancements, GMC and QPel useful too?

And did anyone try to rip that movie on only 1 CD? Because the video bitrate looks like very low... (551kBits/s on a 800Mb CD)

Thanks

[Iznogoud]

Yes, it looks good and you could first try it without the convolution3d. And yes you should use b-frame + GMC, no psycho, no Q-pel.

[haazheel]

Originally posted by Iznogoud
Yes, it looks good and you could first try it without the convolution3d. And yes you should use b-frame + GMC, no psycho, no Q-pel.

And why not psycho or Q-pel?

[Iznogoud]

Q-pel isn´t working wery well but psychovisuals are a matter of taste and maybe you could use them, but they can also make things look worse. Anyways LOTR (atleast the original <3h version) should fit nicely to 2 cds. Best bet: try everything and see what you like:devil:

good luck!

edit: about PE

http://forum.doom9.org/showthread.php?threadid=37762

[?¿öM¿?]

try using BicubicResize(640,272,0,0.5) instead of LanczosResize as your bitrate will probably be around 1100 ? LotR should have good compressability cause of all the dark scenes so maybe LanczosResize will be alright. It isn't recommendable to use a sharpening resizing filter like LanczosResize at a bitrate below 1300 (depending upon compressability). GMC can give a slight compressability gain but sometimes causes visual problems, it is a matter of benefit verses risk. Q-Pel works well for me and if you are going to view it on a TV then Q-Pel definately is recommended. Psychovisual enhancements tend to create a general degredation of image quality from my personal tests, but occassionally on movies which are bright and have lots of fast action it can give a good increase in compressability. Using the above BicubicResize you might also be able to increase your res to around 688xwhatever which gives good fullscreen viewing. You shouldn't really use the convolution and definately use B-Frames, they give a great increase in compressability.
:)

[haazheel]

Originally posted by ?¿öM¿?
You shouldn't really use the convolution and definately use B-Frames, they give a great increase in compressability.
:)

Why not the convolution3d filter?
It give an increase in compressibility too, no?

Or perhaps is it too slow for the poor result in compressibility?

[serbersan]

Yes I think MovieHQ is slow for the result, and you won't gain enough compressibility to use it. And of course you will loose detail if you use it.

The use of a filter is only recommended when you could increase the visual quality of the movie.

In general try to lower resolution before to think in using a denoise filter, or.... use it if you know perfectly what are you doing, like some guys here like Acaila ;)

[Iznogoud]

Originally posted by ?¿öM¿?
It isn't recommendable to use a sharpening resizing filter like LanczosResize at a bitrate below 1300 (depending upon compressability).


Damnit, I have used happily Lanczos with ~700 bitrate and I liked it too! Where did you read this? I was under the impression that Lanczos was THE filter for low bitrates.

[OvERaCiD23]

B-frames only for the Pro options. GMC WILL make panning scenes jerky. I encoded LotR with GMC and panning scenes were not super-smooth; re-encoded w/o GMC and everything was good (not a playback issue, a GMC bug that is well-known). PVE have only degraded quality with no increase in compressiblity in my many tests (have tried on over 15 movies, never a good result). I have not tested QPel much for quality improvements, but compressiblity has never increased in my tests either. So basically, B-frames are the only worthwhile Pro option IMO.

as for compressibility, use GordianKnot's CompTest to check that, but you'll likely need a different resizer (NeutralBicubic most likely, I used Lanczos for a 3 CD rip with AC3 sound ~~~>back in the day, would have only done a 2 CD rip with my current encoding experience).

as for C3D, I like to use it on many of my rips. Using the MovieHQ preset does not degrade quality (visually), and increases rips usually ~6% on the comptest. If you don't mind the extra encoding time, I say it's well worth it. I trust the CompTest over my eyes on issues regrading compressiblity, as I can't see a difference but trust the tools I'm using to evaluate the compression.

so basically, your frameserver is good except for your resizer. try NeutralBicubic and see what it offers. put C3D before your resizing to preserve detail, and run 2%-5% comptests to ensure you have good compressibility.

[Acaila]

Iznogoud wrote:
Damnit, I have used happily Lanczos with ~700 bitrate and I liked it too! Where did you read this? I was under the impression that Lanczos was THE filter for low bitrates.Lanczos was created for Avisynth because it retained so much detail, more so even than Bicubic(x,y,0,0.75). As you might know, the more detail a frame has, the more diffiult it is to compress it. This means that Lanczos is NOT suited for low bitrate video, the various Bicubic flavours are much better for this. If however you have enough bitrate then using Lanczos will give you a better picture, but in general I do not recommend using it for 1 CD rips because the bitrate is usually too low (there are exceptions of course).

[Iznogoud]

Live and learn :confused: -> :D Thanks for clearing that out Acaila, I got my impression from reading (the start of) this thread http://forum.doom9.org/showthread.php?s=&threadid=33290 like weeks ago when Lanczos was new and everybody seemed to like it even for low bitrates. I still like it, maybe because my old telly softens the picture alot. And for LOTR 2CD, I would still atleast try it.

[manono]

Hi-

I've been using LanczosResize a lot, and actively promoting it. But I can only use it less than half the time. I recommend always running a Compression Test to see if the results will allow the use of it.

So, then the question might become, "I get good results in the Compress Test at this resolution with Lanczos, but at the next higher resolution the Compress Test tells me to use Bicubic 0.5. Which resolution and resizer should I use?", or something like that. Personally, I'd go with the higher resolution and Bicubic up to 640*xxx unless the results of the Compress Test says that LanczosResize is OK to use at 640*xxx. But that would be purely a matter of personal preference.

[serbersan]

Hey manono¡¡¡


Have you any proof or test where lanczos was visible better than bicubic resize 0.75. I mean visible.

Only want one example of that. I've been using bicubic(0.75) and lacnzos and never saw where Lanczos is better even with high bitrate I never saw an image where bicubic show artifacts and lanczos no (for example).

Besides it needs more bitrate and using and image to image comparision (in a high quality & bitrate movie)I saw that bicubic retained more detail than lanczos. This efect wasn't visible watching the movie or if you don't see very well both images but was there¡¡¡¡

Sergio

[Iznogoud]

@haazheel

Don´t forget to tell us what you decided to use and the results:)

[?¿öM¿?]

I've noticed that, as serberson said, Lanczos seems to retain about the same amount of detail as Bicubic(x,y,z,0.75) but the image is definately much crisper. I used Lanczos on Blade 2 which has great compressability(very dark) and got a beautifull result. I actually encoded a section in Bicubic(x,y,z,0.75) and again in Bicubic(x,y,z,1) and Lanczos had the same number of objects in each frame, but had significantly sharper lines and edges. Bicubic(x,y,z,1) had a little bit of noise that Lanczos didn't have and still didn't look as crisp. This is the only movie I have done a comparison test with Lanczos on, but now I'm curious and am going to do more on a variety of different movies. ;)

[haazheel]

Hi,

so I've used the following .avs:


LoadPlugin("D:\Tools\GORDIA~1\mpeg2dec3.dll")

mpeg2source("D:\films\DVD\lotr.d2v",cpu=6)

crop(0,65,720,446)

lanczosresize(720,304)


For DivX, I used PE light, B-Frame and QPel...

I got 1041 for the video bitrate.
The sound was a 110Mo ogg file @ 126.

I like the quality of the movie, but sometimes I find that there is a lack of details in some scenes, like when the fire daemon appears in the Moria Mine... But when the movie plays, we don't notice it a lot...

Here are some screenshots from the 2nd CD

http://oppc.free.fr/divers/lotr1.jpg
http://oppc.free.fr/divers/lotr2.jpg
http://oppc.free.fr/divers/lotr3.jpg
http://oppc.free.fr/divers/lotr4.jpg

[JohnMK]

My guess is that the compressibility test would not support your choice in resolution and resizers. Ugly macroblocks.

[?¿öM¿?]

I agree with JohnMK in that due to your bitrate you might have been better off going with Bicubic(x,y,z,0.75) or even Bicubic(x,y,z,0.5) and a lower resolution (688xheight) to reduce the ugly macroblocks and the slight resize would act like a very light noise filter. Your lack of detail is probably partially due to the PE, despite them being set on light they sometimes remove a massive amount of detail from dark scenes that contrary to the DivX Help thing is very noticable by the naked eye. The high resolution probably also means that your avg quantizer is higher than you would like which also removes detail and can aid in creating nasty macroblocks. It can also be a good idea sometimes if you feel brave to increase the max quantizer to something like 16 which gives the encoder a chance to spread the bits much more effectively, especially at such a low bitrate for the resolution and resizer. This can back fire though :eek: .

[DJ Bobo]

I just encoded the US special extended version and wanted to share my results with you guys.

I ripped every DVD on a 90-MIN CD-R, so 1:45:36 on the first CD and 1:37:04 (cut at frame 139639) on the second CD.

I used the original 192kbps AC3 2.0 Stereo Surround track from the DVDs.

So the video was encoded with 830kbps for the first part & 920kbps for the second part.

All DivX settings were left as is, averaging period was changed to 75962 for the first CD, and 69820 for the second CD.
Only Bidirectional Encoding was used (no GMC, no QPel, as always)

I got an average quantizer of 4 for the first part, and 3.9 for the second part, which means the quality is very good (saturation is around 3)

Here is the AVS script I used:

loadplugin("c:\mpeg2dec.dll")
mpeg2source("lotr.d2v")
crop(2,60,716,356)
bicubicresize(640,272,0,0.5)
temporalsmoother(1,1)

I added trim(0,139639) after the mpeg2source line in the script of the second part.

[N_F]

It was almost a month ago that I did this movie. But I think I got a final size of 640x256... Am I alone in this?

[BoNz1]

No, you are not alone, I am doing LOTR as well at 640 by 256. Probably you checked the box in GordianKnot that says follow ITU-R BT .601 Standard which is why it is different, if you uncheck it then it will be 640 by 272. I always thought that you were supposed leave that box checked to get the right aspect ratio at least that is what gknot says beside it.

[N_F]

Well, I decided to see exactly how my avs-script would look (Seems like you have a NTSC version while I have a PAL version):

"ITU-R BT.601 Standard" activated
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
mpeg2source("D:\Temp\fotr.d2v")
crop(6,78,710,420)
BicubicResize(640,256,0,0.5)
Seems to go for 2.5 AR.


"ITU-R BT.601 Standard" deactivated
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
mpeg2source("D:\Temp\fotr.d2v")
crop(6,78,710,420)
BicubicResize(640,272,0,0.5)
Seems to go for 2.35 AR.


2.35 is the number you see at the back of the DVD so I guess that should be the correct one.

Doom9 recommends unchecking "ITU-R BT.601 Standard" in his guide (unless I remember wrong), while TheWef has this to say:

It's recommended to leave this checked.
It is the correct resizing method for DVD.
If you uncheck 'ITU-Standard' the resizing method will be similar to
the regular one used in software DVD-players like WinDVD and PowerDVD
and described in most guides.

But actually this will result in an aspect ratio that is ~2.5% off!
This fact was pointed out to me by jackei, the author of dvd2avi.



TheWef's words are very convincing, but it don't seem to add up (the AR should be 2.35, not 2.5, right?). Oh well, I'm off to do a search on "ITU-R BT.601 Standard" hoping to find some more info and comments.

[N_F]

http://forum.doom9.org/showthread.php?s=&threadid=29181&highlight=ITUR+BT.601+Standard gives some more info on the subject for those who are interested.