Just transcoded Terminator 2:se Got horrid quailty!

[baker]

First off, the author of DVD-RB should be worshipped like a god and handed lots of money!

Next off, I just encoded terminator 2:SE (It comes in a metal box, mines started to rust, but thats anotehr post althoughter ;) )the encoding process went smooth as a childs arse and DVD-RB done a great job sticking all this back together. I used default settings except, I choose to steal 25% of the bitrate from the extras for the main movie.

The quailty of the movie was crapola. What I want to know is why, and also how is DVD-RB choosing its bitrates? The whole idea of encoding only a cell at a time is a new one to me and I would love to know how the bitrate is calculated? Perhaps this is where the problem is, as I have seen scene's which would be difficult for an encoder come out fantastic, and other slow moving scenes come out terrible? This showing perhaps you are doing something like working out the original bitrate on the original DVD, and halving it?
This might not suit CCE and would explain the above quailty issues?

Is Terminator 2 just too much to put on a single DVD-r. Or is there something more sinister at work here?

Thanks!
Baker

[lamster]

Let me start by saying I haven't seen this DVD, so I don't know if the following is relevant.

I saw something similar with X-Men SE; the result was horrible.

I looked at the output from the Prepare phase (which is saved in rebuilder.log if you have "Status Logging On" checked). It showed that it was being compressed down to about 10% of the original size.

There was a DTS audio stream taking up a lot of space, and a bunch of large menus (which aren't compressed by Rebuilder). (To see how much space is being taken up by menus, look at VIDEO_TS.VOB and *_0.VOB.)

Compressing the menus with DVD-Shrink, and having Rebuilder strip out the DTS audio, wound up giving much better results.

[huesage]

@Baker:

If you want some help you should probably give some more details. I think a good place to start would be to include what version of programs specially DVD-RB you are using and also check size of final output.

I for one have been having problems with undersized output using DVD-RB 0.65, but who knows, you have not mentioned any details at all.

Maybe with more information someone will be able to help you.

[Pabloz]

i got the same dvd you have, but original,
ill check the dvd later, but dvd1 doesnt have
much extras, extras are on the second dvd

and yea, this might be a problem with the
menues, it is a great menue plus it has great
audio tracks.

[Pabloz]

one more thing, this movie has like a lot of shit inside
cos i think it has the normal version, the teatrical version
and the extended version with alterate endings

plus its a 2 hour long movie, so you better be carefull

menues could be too big, audio too, and it got like 3 versions of the movie inside

[AzraelsCurse]

Could the quality issue be due to the ILVU portion taking up a lot of space, and not being recompressed? How can we check how much space ILVU takes up? I have this DVD as well but have not tried it yet. I stopped testing multi-angle after Finding Nemo also turned out "horrid" quality wise(w/3 pass CCE). I'm guessing it's the ILVU compression situation.

[lamster]

Originally posted by AzraelsCurse
Could the quality issue be due to the ILVU portion taking up a lot of space, and not being recompressed? How can we check how much space ILVU takes up?
If the DVD does have ILVU, that's certainly a possibility.

One way to check (for someone that has 0.65) is by adding to the Options section of the Rebuilder.ini file:
ILVU_Extract=1

This will generate some VOB's in the D2VAVS directory corresponding with the ILVU / angles. You can see how large these VOBs are. (But see also what I reported about keeping these around causing overflow errors in the 0.65 buffer overflow errors caused by ILVU_Extract-generated VOBs ??? (http://forum.doom9.org/showthread.php?s=&threadid=84384) thread.)

If you don't have 0.65, you could use IfoEdit to open the IFO file for the title set containing the ILVU, and see which program chains are listed as containing ILVU, and add up the times for all cells so marked. This will give you a rough idea of how much it's costing you.
I stopped testing multi-angle after Finding Nemo also turned out "horrid" quality wise
I'm testing everything that I can, because I want to shake out all the bugs now. If the result for a particular movie looks bad, then I analyze what's taking up space, and whether I really want to keep the extra angles, or the features provided by the ILVU. I might decide to strip it, or I might decide that I want it and so I have to wait for a future release that also compresses the angle / ILVU portions.

[AzraelsCurse]

lamster, I added ILVU_Extract=1 to the options seciton in the rebuilder.ini, and then did Finding Nemo once again(I still had it ripped). It produced 0 vob files in the D2VAVS folder. Let me know if there is anything else I need to do to produce these vob files to check the space ILVUs are taking up. If I can get it to work, I will then do the same for Terminator 2 to see if that's the problem. Thanks for the info.

[Pabloz]

http://img59.exs.cx/img59/7206/t15.gif

[lamster]

@AzraelsCurse - Sorry, but that's all that I recall having to do. I'll try it out later tonight.

[pg55555]

Pabloz

You really do not have three independent movies (PCG) in the DVD. You have 3 "chains", wich share more or less footage

From a previous analogy I used:
You have yo go from times Square to Wall Street
- Teatrical release: Straight by Broadway Ave.
- extended version: You go for Bradway but take a 2 block detour in the Village before gong back to Broadway Ave.
- Alternate ending: End in Battery Park

As you can see, most of the road is shared by the 3 routes, but each of them have specific sectors. These are the ones that are addeed to the Teatrical release.

AzraelsCurse
It produced 0 vob files in the D2VAVS folder
You have to complete the encoding and rebuilding process to the .VOB to be generated. They will be in the VIDEO_TS folder and, if I remember, will have "A" and "B" letters in their name (VTS_01A.VOB or something so) and they represent the size of the VOBID that have multiangle/ILVU and are, therefore, not compresed.

[calpchen]

Hrm. I don't know much, but do duplicate PGCs (http://www.doom9.org/index.html?/mpg/Eyes_Duplicate_PGC_Guide_to_Finding_Nemo.htm) have anything to do with the horrible quality on Finding Nemo ?

[Trahald]

Nemo is a duplicate pgc.. its sortof a non-seamless multistory. the way rb processes video the only real factor is total mins video, doesnt matter how many pgcs its in

it doesnt suffer the same issues as a ilvu/angle movie would as there isnt any real duplication of video like there usually is in ilvu.. however the extras and the very large menu it has can lower your available bitrate and be a quality factor.

t2 is ilvu and has duplicated video because of it.. it has about 2hrs 48 mins of video in the movie VTS (actual video) and about 450 mb (again roughly) between the menu and extras.. it doesnt leave much left. the ilvu sections are pretty small ( 13 mins or so estimate) so not compressing them doesnt cause too much impact. if you kill the DTS that helps but you still arent talking alot of leftover bitrate.

[lamster]

Originally posted by AzraelsCurse
lamster, I added ILVU_Extract=1 to the options seciton in the rebuilder.ini, and then did Finding Nemo once again(I still had it ripped). It produced 0 vob files in the D2VAVS folder.
I verified that's all I needed to do to get a VTS_02_A.VOB generated in the D2VAVS directory during the Prepare phase for Star Wars: A New Hope. That has multiple angles; perhaps the ILVU_Extract only works for angles?

[Morbo]

I dont have that disc,but my brother does......and you cant stick the whole first disc on one disc(DVD-5),quality wise......

I think it has a version on each layer,and the original(or visa versa) is an easter egg type button.

Its full of action and extremely long,so I would just do a backup of whichever version you really want to keep...and lose the menus.

I can't wait till my Return of the King comes next month,it should be the same problem...:rolleyes:

Cheers!

[jdobbs]

ILVU_Extract finds any interleaved video. Multiple PGCs in no way implies ILVU or angles. A PGC can just be a different string of cells that are or are not included in another PGC.

I think one quick fix for these high-ILVU discs is the ability to remove angles. I'll probably put that into one of the upcoming releases... the only reason I hesitate is that I really prefer to concentrate on complete disc backups.

[lamster]

Originally posted by jdobbs
I think one quick fix for these high-ILVU discs is the ability to remove angles. I'll probably put that into one of the upcoming releases... the only reason I hesitate is that I really prefer to concentrate on complete disc backups.
You already provide the ability to strip out unwanted languages and subtitles; I see the ability to strip out unwanted angles as an extension of this. (True, angles and ILVU can be used for a lot more, but in many cases they're just used to provide the ability to display written text in video clips in the viewer's preferred language.)

If you're letting me strip out the Japanese audio stream, why not also let me strip out the angle that contains the Japanese text, so I have more room for the video I actually want to see?

Just some food for thought...

[jdobbs]

The quailty of the movie was crapola. What I want to know is why, and also how is DVD-RB choosing its bitrates? The whole idea of encoding only a cell at a time is a new one to me and I would love to know how the bitrate is calculated? Perhaps this is where the problem is, as I have seen scene's which would be difficult for an encoder come out fantastic, and other slow moving scenes come out terrible? This showing perhaps you are doing something like working out the original bitrate on the original DVD, and halving it? Sigh... how many times do I have to explain this? Not again. It works, leave it at that. If you don't like how it works -- use something else -- or better yet, write something yourself like I did. I'm getting pretty tired of this constant barrage of pseudo experts lobbing hand granades from left field.

[baker]

quote-
Sigh... how many times do I have to explain this? Not again. It works, leave it at that. If you don't like how it works -- use something else -- or better yet, write something yourself like I did. I'm getting pretty tired of this constant barrage of pseudo experts lobbing hand granades from left field.
Unquote,

I am not even close to being an expert, in fact I am an incredibly stupid person! And I will readily admit this! I havent wrote anything myself as I don't have the time and wouldn't have a clue how, besides, I already have DVD-RB ;)

I am using dvd-rb as I want to keep the entire dvd, even feckin' arabian subtitles! Its not that I am ever likely to be able to speak such a language, its just that I want to keep the WHOLE dvd.

I just wanted to know how it calculates its bitrates for each cell, purely out of interest, I would like to know how it accomplishs this and if it's already posted somewhere, a link would be greatly appreciated!

Baker

[pg55555]

baker
I just wanted to know how it calculates its bitrates for each cell, purely out of interest, I would like to know how it accomplishs this and if it's already posted somewhere, a link would be greatly appreciated!
I'm not an expert, but from my understanding of previous answers by jdobbs on the subject (more or less):
- The digitalization and encoding of the original movie (the commercial DVD you bought) already did the allocation of the bitrate: remember, the DVD is already compressed mpeg2 video and AC3 audio
- So, RB, by dividing the video in segments, uses this pre-existing bitrate allocation, as it would be a kind of "first pass".

Regarding your original posting, huesage (in the third post of this thread) suggested a way to find a solution to your problem

[baker]

Just like I tought!

Thanks!!

Perhaps my idea was right on the bitrate selection.... scenes with highier original bitrates where comming out better as CCE still had a decent bitrate to work with, but scenes that where originally encoded with quite a low bitrate, are been given too little of a bitrate. Thus, high action cells end up comming out better than low action ones!

I could be wrong on this, so I am going to check out some picture distortion rates on that T2 dvd I transcoded. I'll post the results for those of you that are interested, when I get time.

Baker

[jdobbs]

One thing to remember is that there are some DVDs that are already stretched to their limit in terms of bandwidth -- and no amount of additional compression will result in acceptable quality.

I heard an example the other day of someone who was trying to take an original DVD-9 that had 5 hours of video on it and than compress it even further to put it on a DVD-5. That would result in a video bitrate of about 1.8Mbs even if you only had a single 2 channel audio stream... and no menus. That's a real stretch for 720x480 no matter what encoder you use. Even CCE has a hard time with that kind of MPEG-2 demand.

Remember that sometimes even the best you can get isn't good enough.

[pg55555]

baker

I do not have T2 SE, and as you have not post any detail regarding the DVD structure (except for the fancy packaging)I can only guess.

If T2 SE has, as other SE, two angles / ILVU chains, one with the Theatrical release and other with the extended release with deleted scenes, the following can be happening:
As the angles/ILVU VOBIDs are not being compressed, there is less space for the ones that are being compressed.
So you have slow motion scenes with "high bitrate", because they have actually not been compressed.
And high action scenes with "low bitrate" wich have been compressed in excess to make room for the uncompressed parts

[baker]

Well pgc55555!
Nice idea, but that isnt correct.

I'm afraid at the end of the day, the edition of T2 that I have is simply too packed. Even if everything (including menu's etc) was compressed, the quailty still wouldn't be that good, its simply too much for a single dvd-r!

My point on using the original bitrates as a guide was correct, however there isn't enough difference, especially on an ordinary dvd, to warrent a change in the DVD-RB coding. Theres still plenty more things to be done before we start nit-picking over a few kbs ;)

I think we'll finish the thread here! Sorry for not posting more on the T2 DVD in the first place, but I hadn't the time!

I just wanna say best of Luck to you with DVD-RB, and I hope you make a few bob out of it!

Baker

[jdobbs]

Cheers.

[TheSeeker]

@ Baker

Sorry to revive a dead horse and then proceed to beat the crap out of it, but actually pg55555 actually has a very good point. That is most likely what is happening. As you may have noticed in another thread that the multiangle version of RB do not compress any ILVU ID's. and Since T2: UE is most certainly multiangle, and just listening to what you were saying (that certain parts looked great, and some looked horrid) I would say this is exactly what is happening. I find it just a bit annoying that you dismissed pg55555's suggetion immediately. I feel it to be the cause of your problem as it makes the most sense.>. I believe jdobbs is addressing this issue so if your patient then you may be able to use RB to do this disc. I dont feel that it is too much to pack onto a single dvd-r and get decent quality. Especially if you dont mind getting rid of an audio track or two.

[baker]

Sorry, but pgc55555 was WRONG! It was a correct idea, and could happen. But thats not what it was in this case.

Also, losing an audio track or two, is NOT the full dvd ;)

quote- I find it just a bit annoying that you dismissed pg55555's suggetion immediately.
Unquote.

I did not "dismiss his idea immediatly" I simply said it was wrong!! :b

Pgc55555 didn't have the full idea of what was on the DVD, so his answer "could" have been correct. However, as I have the DVD I know its not!



Baker

[TheSeeker]

Ok, I can accept that... Then what is your explanation for the quality being really quite good, then horrid, then good, etc. You have to admit it sounds like exactly what pgc55555 said... But if you know what was the matter than I would like to hear it. Just so I know because I will probably want to encode this dvd myself. Oh and just so I know... Do you speak all the languages that are included in the audio tracks of the dvd?

[baker]

quote=Then what is your explanation for the quality being really quite good, then horrid, then good, etc. You have to admit it sounds like exactly what pgc55555 said... But if you know what was the matter than I would like to hear it. Just so I know because I will probably want to encode this dvd myself. Oh and just so I know... Do you speak all the languages that are included in the audio tracks of the dvd?
/quote

Its got to do with DVD-RB using the original bitrate of the DVD as a guide.

As pgc55555 said-
- The digitalization and encoding of the original movie (the commercial DVD you bought) already did the allocation of the bitrate: remember, the DVD is already compressed mpeg2 video and AC3 audio
- So, RB, by dividing the video in segments, uses this pre-existing bitrate allocation, as it would be a kind of "first pass".


And as I said-
I am using dvd-rb as I want to keep the entire dvd, even feckin' arabian subtitles! Its not that I am ever likely to be able to speak such a language, its just that I want to keep the WHOLE dvd.

Baker

[TheSeeker]

Ok that is another possible explanation... However when I did the movie The Last Samurai, it came out fantastic. I didnt get rid of anything and the movie was really big, and also quite long. A little longer than T2 I believe. It also has high action scenes of big battles. I saw no evidence of high action scenes good, low action bad. I have done many movies with rebuilder that have high action scenes and that are being compressed ALOT (sometimes as much as to 55% of original) and I dont see this behavior. So dont you feel that either one of us COULD be right? I mean there is strong evidence for both cases... I wonder if the man himself Jdobbs could shed some light on this for us?? I just dont like how you feel that your right and everyone else is wrong when you have to admit that their is strong evidence for either case...

EDIT: Oh and your right it does use the original bitrate as a sort of guide... but only in a very loose sense.. I mean that IS what the first .vaf pass is for. to determine where more bitrate is needed. I think if the first pass found the low action scenes to be of such a low bitrate that it would produce horrid quality it would assign a little more bitrate to those scenes dont you? CCE is quite an intelligent encoder after all.

[jdobbs]

The original DVD is used as a guide as to how much of the entire bandwidth available is allocated to an individual cell. Then VBR is applied within that cell to determine how much of the allocated bandwidth is applied against each scene/frame.

After many, many tests I have concluded that it is every bit as accurate at predicting correct bitrates at the scene level as when using a single VBR domain against the entire movie.

What Baker is probably seeing is that the "high action" scenes may also be the ones that are ILVU. So they weren't reencoded -- and as a result appear to be clearer. But -- that isn't a sure thing either, because ILVU originals tend to be more tightly encoded than the rest of the movie (because they have to be interleaved).

[pg55555]

TheSeeker

Thanks for your support, but as I have told in my post, I was GUESSING, as I do not have the original DVD and baker , who has it, has not posted any detail about its structure, except the size.

So the probabilities of me being WRONG were high. And I have to believe I was.:(

But as baker choose to not provide the extra information and to give up, his problem!:D


Other point: sometimes we talk about "low" and "high" without really knowing what it means. RBOpt would provide us with the info about the average bitrate calculated for each segment or cell, and tweak it if we like. And PowerDVD's can display the bitrate of the frames when viewing the movie. So maybe in the next analysis/discussions we can be more specific

[jdobbs]

"Dynamically Assign Bitrates" is what I was discussing. If you don't select dynamic allocation (the default) -- an average bitrate is calculated using the entire group of video and is applied against it, and the allocation (between cells) of the original is not taken into consideration.

[baker]

I am not sure how to explain what is going on, I have it in my head, but typing it here is difficult!

In any case it only matters where the bitrate was too low in the first place anyway and the difference is miniscule! (is that even nearly spent right!?) A couple of kbs!

IT IS NOT AN ILVU PROBLEM!!!!! PLEASE UNDERSTAND THIS!!!! IF I GET ONE THING THROUGH, PLEASE LET IT BE THAT!!! :)

I'll post a few screen shots to prove this, and if i get a bit of time, I will show the cause of this situation in detial and explain how it happens.

But basically, "cells" which have an originally highier bitrate (due to the way they were encoded originally, with a much highier ABR), are given a highier bitrate by DVD-RB. Cells that are encoded with a lower bitrate are given an even LOWER bitrate, causing low action scenes to look worse and high action scenes to look better!

Its like this. Take two scenes in film Z, each lasting 10 secs.

Scene y has loads of action and due to having a ton of bitrate as it was a DVD-9 they where authoring, it got a average bitrate of 5000.

Scene x has little action and is just 2 people talking about something. It only got a bitrate of 1800 (ample for the scene).

Now, if those two scenes where cells in a DVD that DVD-RB was encoding, it would (and it dosn't actually do this, but this will make it easier to explain) Half the bitrate of each cell.

Now scene y has got a bitrate of 2500, the scene looks ok as CCE can handle action scenes with an average bitrate of this ammount.

Now scene x has only got a bitrate of 900!!! Its a low action scene, yet looks tripe as the Cell was given too low an ammount to cope with the scene (even though it was only low action.)

Now do you see where the problem is!
But its a very rare thing to happen, and Film X was an extreme case. In real life, its not likely to matter that much.

P.S. Dynamically assigning bitrates causes this, but is a lot better than simply applying the average bitrate to each cell! So until DVD-RB has got to the stage where we can start looking for ad-ons, it will have to do.

Baker

[TheSeeker]

Ok, so then my question for you is this.... Why not use RB-Opt to assign a little bit higher bitrate to those low motion scenes that arent getting enough bitrate? It allows you to assign bitrate on a per cell level so it should be exactly what you need to fix this issue. Have you looked into it at all? Also you could apply the kvcd notch custom matrice which works wonders with really low bitrate material. You should check it out if you havent already.

EDIT: I just wanted to say first off I believe you.. your theory is probably right so im not starting this arguement again I just want to pose the question this whole thread planted in my mind. Since this is caused by RB reading off from the original dvd's bitrates then how is this NOT happening more often. I mean there must be alot of movies that are long, take up most of the dvd and have plenty of low action scenes. And really those are the ingredients that cause this problem arent they? Or is it that that disc was so poorly authored?

[baker]

Well if you think about it, it actually happened cause the disc was WELL authored!

Don't just believe me by the way. I could be wrong. (And I am waiting for someone to prove to me that I am wrong as I well could be, and if I am not then I would like to be proved corret!) Just cause it makes sense dosn't mean that I am right, the earth isn't flat!

Anyway, if I am right, then the reason this is not happening more often is as the bitrate is never as "diverse" between the scenes as it is in film X. Plus the difference in the quailty of the different cells in T2 wasn't enormus, however it was enough for me to notice.

The fact is, if I removed the DTS stream and a few others, the bitrate would be increased and this issue might not be apparent at all.

Baker

[baker]

Well if you think about it, it actually happened cause the disc was WELL authored!

Don't just believe me by the way. I could be wrong. (And I am waiting for someone to prove to me that I am wrong as I well could be, and if I am not then I would like to be proved corret!) Just cause it makes sense dosn't mean that I am right, the earth isn't flat!

Anyway, if I am right, then the reason this is not happening more often is as the bitrate is never as "diverse" between the scenes as it is in film X. Plus the difference in the quailty of the different cells in T2 wasn't enormus, however it was enough for me to notice.

The fact is, if I removed the DTS stream and a few others, the bitrate would be increased and this issue might not be apparent at all.


Oh yes, and I could use rb-opt, but the result would most likely be that ALL scenes look crap :)

Baker

[TheSeeker]

Possibly... But you dont know until you try right? But it sounds like youve given up on dvd rebuilder for this film anyways... Im just going to have to get it and give it a whirl... Just wondering though, is completeness really more important to you than quality? Or is it just this movie that you are so adament about not removing anything? Because you cant always have both. Actually usually you cant have both. If you want really good quality on a dvd-r you usually have to get rid of something. Not dissing you at all just curious.