"bad quality" even with 2 cd rip

[volkl23]

Hi Everyone, I have a major problem that's been giving me a major headache. I tried searching through the forum for my particular problem, but none of them address them well.

Here is my situation...

I've been trying to encode various movies ("1o things i hate about you," "A Knight's Tale," and "Don't say a word.") using Divx 4.12 and Gordian knight with a 2 cd rip.

Here is my info for Knight's tale..

Video: 592 x 240 using .392 bits/pixel and a compress of 54%.
I don't have the log with me (I trashed it), but my bitrate came out too approximately 1300-1330 for a 2 cd rip.

Audio: using 128k VBR

I also used sharp bicubic with no other filters.

When I played it, I wasn't happy with the quality at all. Overall the movie has really really small blocks. I don't know if these are macroblocks, because they're so small, but it's essentially the same as when you zoom in on a jpg too much. That "pixelated effect."

I just don't know why it's so crappy with 2 cds. On another note, I might add that the movie looks just like the dvd at my original encoding resolution (592x240), but looks like crap (lots of really small blocks) when viewed full screen. When people say that their rips are really good, are they usually good full screen or just at the native ripped resolution. I'm asking this, because my rips might be "good" after all then if other peoples rips are just as bad when viewed full screen.

My goal is to just remove those darn small blocks. I tried using soft and neutral bicubic (as opposed to sharp bicubic), but the tiny pixelation blocks are still there when viewed full screen. They actually look more worse than sharp bicubic in my opinion.

I've been told that "A knight's tale" is a bitch to rip, because it's very uncompressible (i only got 54% using 592 x 240 with a 2 cd rip!!), but I have tried ripping other movies that are slightly better at compressing, but I still get those darn blocks when viewed full screen! My screen res is (1024x768 btw).

I was using Flask .594 before using Gordian, but I've read at a lot of places that flask is not that great and that Gordian is a lot better, so I switched. Now going back to the movies i ripped using flask, they actually look pretty good! The small blocks are somewhat noticable, but the resolution is generally set a lot higher as well. Most of the movies i did with flask are at around 704xxx, so I'm guess that if I reduced the resolution to what Ive been using with Gordian Knight (through compressibility tests), the blocks might be gone althoghther! What are your thoughts on flask vs. Gordian knight?

What am I doing wrong?? :(

[Enchanter]

Hmm...that's a pretty hard one to solve. Generally, at a bitrate this high, you shouldn't get into much trouble. Heck! You could actually use a higher resolution like 640x256 and still stand a small chance of getting macroblocks.

Try using bilinear instead. It may blur the picture, but that also makes the picture more easily compressed.

In addition, you could try using the DVD2AVI project -> vFAPI method. This is the method I've been using for quite some time. I recently started using Avisynth as well, but somehow the results look slightly 'not as good' (Nowhere near bad) as the vFAPI method. Could just be my feelings though, but you could give it a try as well. :)

Lastly, I would still stick with GK+Nandub, rather than Flask. The former wins hands down in the number of encoding options, quality, etc. :)

p.s. Why don't you share your encoding methods here and we'll have a look if there's anything wrong.

[khp]

Video: 592 x 240 using .392 bits/pixel and a compress of 54%.
I don't have the log with me (I trashed it), but my bitrate came out too approximately 1300-1330 for a 2 cd rip.


At that bits/pixel value the video should be very close to beeing saturated, at least for '10 I hate about you', which I have done with divx4.12 in GKnot at 704*X and it came out perfect.

your problems probably is that the movie is not displaying at the proper color depth. Try switching to 24 or 32 bit in screen->properties->settings', if you are using 32 try switching down to 24.
The problem might be caused by vdub setting the colordepth flag incorrectly in the avi file when using fast recompress (which is what GKnot does).

If all else fail try loading the avs file created by gknot into vdub and convert that using full processing mode.

[Enchanter]

The problem might be caused by vdub setting the colordepth flag incorrectly in the avi file when using fast recompress (which is what GKnot does).

Now you got my attention. :) What do I do to avoid this? I just started using GK as of today and so bear with me. :) I was thinking of using GK to write an .avs file, but basically only for framerserving and IVTC.

[volkl23]

At that bits/pixel value the video should be very close to beeing saturated, at least for '10 I hate about you', which I have done with divx4.12 in GKnot at 704*X and it came out perfect.

Boy, I wish I could get a good rip of "1o things i hate about you." I did many countless hours of tests on this particular movie using different resolutions and different resizing filters.

The best I could come up with was 640x336. Again, it looked perfect at this size, but when i full screened it, the annoying little tiny blocks appeared.

I tried doing the different color depths, but it didn't solve it.

I remeber I had downloaded a 1 cd rip of this movie (704x432) a few months ago. Overall, it was worse than my 2 cd rip. But here's the strange thing.

Although the 1 cd rip I dled had many many really really big macroblocks when there were a lot of motion (as expected with a 1 cd rip), the edges were rather sharp. What I mean by this is that there were far less little annoying tiny blocks found compared to my 2 cd rip. Isn't that weird? Minus the big macroblocks of the 1 cd rip, there were hardly any (there were still some) of the really tiny "pixelation" blocks found.

Does this have anything to do with me resizing the movie to full screen mode?

[techno]

Just out of curiousity, have u ripped the DVD using DIVX 3.11alpha Low motion?

What quality/size do you get? No harm in trying or experimenting.

:)

Techno

[volkl23]

How do you use low motion with Gordian Knight (nandub + 3.11 alpha).

I don't think there is a setting that let's you choose fast motion and low motion methinks.

[techno]

Ok,

in nandub, use the MSMPEG4 v2 codec

Or use other DVD rippers which allow low and fast motion codecs to be chosen.

www.divx-digest.com/software

Techno

[volkl23]

Hi techno, I don't know how to use Nandub all that well. I don't think I even know how to do what your asking with nandub. What ripper that's capapble do you recommend (there's so many out there)? Flask?

Thanks,
volkl23

[techno]

Well, I don't rip dvd's but I have a fair Idea. UnCasMS is very knowledgable and so is enchanter.

Use one of these software to rip dvd's (www.divx-digest.com/software):

DIVX Machine
DVD universe MPEG4 packages
EasyDIVX
RipitAll

personally, the second and first seems good.

remember to install the DIVX 3.11alpha codec.

Let me know what happens

Techno

[khp]

Originally posted by volkl23

The best I could come up with was 640x336. Again, it looked perfect at this size, but when i full screened it, the annoying little tiny blocks appeared.


OK, so the problem only shows up when you resized to fullscreen durring playback ?, you might try using the bicubic resize Direct show filter, you can find this in the download section at www.doom9.net , but be aware that this is a real hog for cpu power.

Originally posted by Enchanter
Now you got my attention. What do I do to avoid this? I just started using GK as of today and so bear with me. I was thinking of using GK to write an .avs file, but basically only for framerserving and IVTC.

There is absolutly nothing wrong with the avi file other than a flag in the header that is not properly set. Most people has no problems with this, a few days ago someone reported a simmilar problem at doom9's forum, he solved it by switching from 24 to 32 bit color.

[techno]

No harm in trying different methods, is there?

Techno

[khp]

Originally posted by techno
No harm in trying different methods, is there?

Techno

No, in fact I always encurage people to try several different methods before deciding which they like best.

But I also try to find the actual cause of the problem, as opposed to your usual 'blame it on divx4' approach. If in fact volkl23's problems are caused by resizeing durring playback, there is not much chance that divx3.11 or any other codec would solve the problem.

[techno]

We shall see about that

Techno

[Enchanter]

There is absolutly nothing wrong with the avi file other than a flag in the header that is not properly set. Most people has no problems with this, a few days ago someone reported a simmilar problem at doom9's forum, he solved it by switching from 24 to 32 bit color.

I'm not sure whether this is a related problem though, but here it goes. I did a 20-min episode using the DVD2AVI -> vFAPI project, and more recently using GK -> .avs file. Both methods used nandub for the final encoding. The same encoding settings were used and the final filesize were close (200KB+ difference, which really is negligible). However, the vFAPI method resulting file looks slightly better, as in clearer and 'less-pixelated' black areas and other small details. I thought this had something to do with the GK settings, but beats me which one. Both methods were using bicubic (vFAPI) and Neutral bicubic (which should be more or less the same with the normal bicubic) respectively. IVTC were performed, with vFAPI using TMPGEnc and GK using its built-in IVTC capability.

Do you have an idea how this can come to be? Both results were excellent anyway, just that the vFAPI-framerserver gave slightly better result.

[khp]

Originally posted by Enchanter

Do you have an idea how this can come to be? Both results were excellent anyway, just that the vFAPI-framerserver gave slightly better result.

Sorry, I can't offer much of an explanation there, problems with the colorspace will be more than slightly noticable, so I don't think this is your problem.

There are a number of differences between using vfapi and avisynth, so slight difference in the output is to be expected, but I doubt that one is consistently better than the other.

[volkl23]

Hi Techno, I actually did use 3.11a with Gordian Knot as well. This was using the 2 pass technique. Overall, I still get that annoying p
ixelation effect when viewed full screen. Overall, I liked the 4.12 just a little bit, but that's not saying much.

OK, so the problem only shows up when you resized to fullscreen durring playback ?

I wonder why I get this when viewing full screen. Well, I just finished a 2 cd rip of "Don't Say a Word." I must say that I have never encoutered a movie as compressible as this. This was definetely on the other extreme end of compressibility. Even at a resolution of 704x288, I had a compressibility of 113%!!!! I a compress of this high is not that good, so I tried increasing the resolution, but I started to go over 100% W-zoom and I heard that's even worse. Do you guys agree with this and what would you have done in this case with the resolution setting? I used sharp bicubic by the way.

Results:

The overall file came in at around 1.09 Gb, which is rather small. I guess I can increase the bitrate setting a little higher to use the slack.

Well, the quality was pretty good even when viewed full screen. There were still a lit bit of that pixelation effect around contrasting edges, but this was less than the highly uncompressible Knight's tale.

I am going to rip 10 things i hate about you and include some screenshots containing this pixelation I'm telling you about.

Before I do so, I want to get everyone's own personal opinion. Is there an ideal compressibility number we should strive for? In the Gordian guide, it said aim for 45-55% while I've heard others say they won't do anything below 60%

Another related question is this.. how far would you trim the resolution in order to get these ideal compress numbers? Is it too bad to go below a certain resolution, even if doing so is the only way to reach 45-55% compress? Some of these numbers I've heard going around are 512xxx and some higher. Your thoughts on this?

you might try using the bicubic resize Direct show filter, you can find this in the download section at www.doom9.net , but be aware that this is a real hog for cpu power.

Can you tell me more about this filter? I tried searching for it at Doom 9 under the dls section, but was uable to locate it. I would rather make a movie that didn't require this- I want to make a movie as stand-alone as possible, but I will use it if it's the only way to get decent results.

thanks again everyone,
volkl23

[techno]

hmmm

weird, completly weird.

Maybe it has to do something with the DVD encryption? (just a guess)

Techno

[khp]

Originally posted by volkl23
I am going to rip 10 things i hate about you and include some screenshots containing this pixelation I'm telling you about.


Great, if you can also note the frame numbers, or at least the time of the image, I will be happy to compare them with my rip.

Originally posted by volkl23
Before I do so, I want to get everyone's own personal opinion. Is there an ideal compressibility number we should strive for? In the Gordian guide, it said aim for 45-55% while I've heard others say they won't do anything below 60%

Another related question is this.. how far would you trim the resolution in order to get these ideal compress numbers? Is it too bad to go below a certain resolution, even if doing so is the only way to reach 45-55% compress? Some of these numbers I've heard going around are 512xxx and some higher. Your thoughts on this?
[/B]

Sorry I never use compressability check, I'll just go by bits/pixel numbers and my general impression of the movie. For very long uncompressable movies (like pearl harbor) I'll go for a 3 CD rip.
If I get a movie that is too compressable to fill 2 cds at full res
I would probably resize to 640*X or 576*X and go for a one cd rip.

Originally posted by volkl23
Can you tell me more about this filter? I tried searching for it at Doom 9 under the dls section, but was uable to locate it. I would rather make a movie that didn't require this- I want to make a movie as stand-alone as possible, but I will use it if it's the only way to get decent results.
volkl23 [/B]

In the dl section at doom9's scroll down to the filters section, and click 'show me more' the Direct Show filter is the second on the list.

Originally posted by techno

hmmm

weird, completly weird.

Maybe it has to do something with the DVD encryption? (just a guess)

Techno


Why do you think this is wierd ?, did you seriously think this was a codec problem ?, what made you think that ?

The fact is that there are very few problems that are truely codec related. One of the few exceptions is divx3.11 producing ****frames. As descriped in this thread (http://forum.digital-digest.com/showthread.php?s=&threadid=4991) which is why I never use divx3.11.

[volkl23]

Great, if you can also note the frame numbers, or at least the time of the image, I will be happy to compare them with my rip.

Cool! That would be nice of you! I'm curious to see other people's rips of the same movies :)



Sorry I never use compressability check, I'll just go by bits/pixel numbers and my general impression of the movie. For very long uncompressable movies (like pearl harbor) I'll go for a 3 CD rip.


What sort of bits/pixel number do you usually strive for in a 2 cd rip? How many rips do you usually do on average before you like the quality of the movie? Just curious.

[khp]

Originally posted by volkl23

What sort of bits/pixel number do you usually strive for in a 2 cd rip? How many rips do you usually do on average before you like the quality of the movie? Just curious.

I usually try to stay just above 0.2 bits/pixel, but you have to allow for ajustments, depending on the amount of action in the movie.
I generally only rip the same movie once. But I also do a lot of testing, to figure out which codec looks better and which set of features works best. Some movies I have done more than 15 times.
Actually I have not done any rips, that I plan to keep, in the last 2 months.

[volkl23]

I usually try to stay just above 0.2 bits/pixel, but you have to allow for ajustments, depending on the amount of action in the movie.

Do you even pay attention to the compress then? I use Gordian and a lot of times, I can't reach .25 let alone .20 without either getting a really horrible compress (like .30-.35) or a resolution that is more than 100% of the original.

For example with "10 things i hate about you," at bits/pixel of .299, I get a resolution of 704x384, W-zoom of 99%, but an H-zoom of 111% which is a lot higher than the original resolution.

I've read from other people that they only look at compressibility and not bits/pixel. Your view seems to be on the other extreme.

Now I'm totally confused :confused:

When you test.. how do you usually go about it? This is what I do and I think it's flawed, but I don't know anyother way to do it.

Basically, I first get the avg. bitrate for the movie for a 2 cd rip (1600kb for example).

I then go to dvd2avi and create a short clip (10 minutes or so that includes many different types of scenes). I input the average bitrate I got from the whole movie and use it for the short clip.

Now comes the part where I think my methods are flawed. When I do a compress check for the shortened movie, it is usually drastically different then from the whole movie. For example .55 vs .40.

I basically run different resolution settings and resizing filters on the different short clips to find the "optimal" settings, but this is sort of wrong, because I usually get different qualities with the final movie, because the compress was so different.

Can you khp or anyone else clue me in on a better way to "test" settings?

Thanks again everyone :)

[khp]

The problem with the compressability check, is that it only tests a fraction of the movie (10% or so). Ofcause trusting bits/pixel values is not enough either, I do make adjustment depending on the amount of action, but how much is always a guess.

When I do test it's usually to test the features that impact compressability, like how much you gain by chopping off the black bars, how much it costs to use imbedded subtitles, what kind of effects the new features in divx5 has, and so on.
The way I do this is that I use 1-pass quality based encoding (at max quality), and use filesize as a measure for how well it works.

[techno]

KHP, u should have better knowldge on DIVX 3.11alpha. Why do u insult everyone that is trying to help????

What I meant by weird was that maybe the DVD disc has been encrypted in a special way, as we all know, DVD ripping is illegal.

Techno

[khp]

Originally posted by techno
KHP, u should have better knowldge on DIVX 3.11alpha. Why do u insult everyone that is trying to help????


I do have quite an extensive knowledge of divx3.11.

And I don't insult everyone who tries to help. I very much prefer to let people make the their own minds about which codec to use. That means that I really hate to advice people to use one codec instead of another, unless it really is the codec that is at fault.

My point is that all you advice, people using anything but divx3.11, is to use divx3.11, regardless of what the problem is.
If you don't know what the problem is it's kind of silly to just blame divx4. I am sorry if I have become a bit rough with you, but telling people doing DVD rips to use divx3.11 fast motion is rather silly, and you have done so on numerus occations, maybe you were thinking about TV captures, but you completly fail to make that distinction clear.