Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > (Auto) Gordian Knot

Reply
 
Thread Tools Search this Thread Display Modes
Old 19th November 2002, 07:43   #1  |  Link
ScarMouth
Registered User
 
Join Date: Jul 2002
Posts: 27
Create a DVD2AVI project - Step 5: Select a resolution

Hello all.

In step 5 specfically,
Quote:
Now you have to set the resolution by moving the big slider.
Move the slider to where? I mean, is there a general rule to the resolution?

The DIVX 4 example show a resolution of 608x320
The DIVX 5 example show a resolution of 640x272

I see how these vales are affected by CROP hence, the different resolution stated above as per the movie backed up during creation of the respective examples.

However, is there a reason to muck with the slider?
Especially if you only work with hollywood stuff?



Last edited by ScarMouth; 19th November 2002 at 07:54.
ScarMouth is offline   Reply With Quote
Old 19th November 2002, 09:25   #2  |  Link
JohnMK
Registered User
 
Join Date: Sep 2002
Location: Seattle
Posts: 551
Yes. Rule: run 5% compressibility test, then choose the resolution that gets you to 60% or so. Don't go above or below this figure, generally.

Alternatively, aim for .200 or higher Bits/Pixel*Frame.

The former strongly recommended. The latter only recommended if you're in extreme haste to pump a movie out. Have no fear about choosing resolutions of 528x224, 576x256, etc. There is nothing holy about having a resolution that has nice pretty numbers we're all familiar with, like 640 or 720 or 512 or whatever.
JohnMK is offline   Reply With Quote
Old 19th November 2002, 17:01   #3  |  Link
jggimi
Moderator Emeritus
 
jggimi's Avatar
 
Join Date: Dec 2001
Location: The NW corner of Lake Erie
Posts: 5,552
The general recommendation is to use b/p*f to estimate size prior to your first compressibility test.
jggimi is offline   Reply With Quote
Old 20th November 2002, 02:34   #4  |  Link
JohnMK
Registered User
 
Join Date: Sep 2002
Location: Seattle
Posts: 551
Good point.
JohnMK is offline   Reply With Quote
Old 20th November 2002, 06:00   #5  |  Link
ScarMouth
Registered User
 
Join Date: Jul 2002
Posts: 27
Thanks for the reply guys...

I thanks you for the info.
However I'm still not into it I guess
This is because my WILLOW rip keeps giving me a frames too long
and not tall enough...

Please bear with me as I walk through the steps. I'm also resetting my config back to default.

1) rip with Smartripper.
2) DVD2AVI ASPECT 16:9, 29.970 FILM.
---> VOB are 720x480
3) SET FORCE FILM. save DVD2AVI project.
4) Open DV2 file.
5) Set DIVX 5 codec.
---> I dont care about number of disks as I'll drop the AVI's on a DVD.
---> Setting b/p*f to at least .200 (.250?) assumes 128 VBR?
---> So if I were to use 160 VBR, I'd do a higher b/p*f?
---> Anyway, I'll use 128 VBR (for now)...
---> I know, I have a DVD burner, why mess with the knot?
---> I haven't found an *affordable* MAESTRO yet.
---> and, processing on main PC is a drag.
---> However, just wired a second box to my PC.
---> Am now considering the MAESTRO path
6) Select 1xVBR mp3 @ 128 kps
---> Input resolution defaulted to NTSC (720x480). Correct!
7) AUTOCROP then SMARTCROP
8) INPUT ASPECT RATIO NTSC 16:9
9) Resize DVD2AVI window...I suppose the frame is OK...
---> At this point I'd mess with the slider bit first, I should
---> set a file size since I wasn't concerned with it before.
---> I going to set a 900meg 1 CD image that produces a b/p*f of .265
---> My optupt resolution defaults to 640x272 with a 0.0 aspect ratio.
---> Great eh?
---> BTW, slipping subtitles.
---> BTW, I started Windows Media player and compared DVD2AVI RESIZE
---> to the actual DVD. Seems like both aspect are inline
10) Preview window SAVE & ENCODE.
---> Neutral Bicbic.
11) Compression test @ 5.
---> b/p*f .265, 84%, .316 compress test
---> Jumping to a resolution of 704x288 changes as follows:
---> ---> b/p*f .228, 72.2%, .316 compress test. Aspect error -0.1 !
---> ---> Jumping any higher and width goes over original 720!
12) SAVE & ENCODE!
---> bi-directional enc
---> gmc
---> psychovisual: light
---> recalc bitrate if needed.
---> MP3 expert @ 160
---> -L -3db -c normal (lame & besweet)
13) Start encode!

Geez!
How'd I do?
oh yeah, didn't scale the cresits down becasue they were mixed into the movie (instead of starting on a black background).

Oh yeah again, since I'm going to be dropping on a DVD with other flicks, should I increase the file size or is the b/p*f @ .228 ok?

Thanks!

Last edited by ScarMouth; 20th November 2002 at 06:17.
ScarMouth is offline   Reply With Quote
Old 20th November 2002, 08:04   #6  |  Link
jggimi
Moderator Emeritus
 
jggimi's Avatar
 
Join Date: Dec 2001
Location: The NW corner of Lake Erie
Posts: 5,552
Hmmm...

"I dont care about number of disks as I'll drop the AVI's on a DVD."

If you're following Doom9's guide, you are attempting to create an .avi file of a specific size. The two-pass encoding method is designed to create the best possible video output of a chosen size.

You could take the .avs script produced by Gknot and do a single pass encoding in Virtual Dub manually, setting the quality to 100%. But that wouldn't be following the guide.

I recommend that you try to follow the guide step-by-step for your first few encodings. Gordian Knot is an incredibly flexible tool, and allows for all sorts of manipulations by the operator. That great flexibility makes for great complexity. Try sticking with a 1 or 2-CD rip for now, and follow the guide. You'll be able to skip splitting any resulting 1.4GB .avi once you're done, of course.

After you get some familiarity with GKnot, and the Vdub/Nandub tools, you'll be able to save some time by doing a single-pass encoding. But I recommend doing that only after you've had a few successful encodings, first.

"Setting b/p*f to at least .200 (.250?) assumes 128 VBR?"

No. It's video only. But it uses the calculated bitrate, and of course that's impacted by the desired final .avi size, and any sizes (estimated or actual) in Audio A, Audio B, or Files.

Don't get hung up on b/p*f. Use it only as an estimation for a compression check. After you get the results from the compression check, use those results, and ignore b/p*f.

"...I have a DVD burner, why mess with the knot?"

Once you've done a few of these per-the-guide, you'll still want to use GKnot -- 1) You'll still it as your AviSynth (.avs) script creator. You'll be editing the AviSynth scripts manually afterwards, perhaps, but you'll still use it as a script generation tool. 2) You'll use it to examine individual video frames when you're working with the unusual NTSC dvds that aren't from video camera sources, but still can't be Force FILMed.

"Resize DVD2AVI window...I suppose the frame is OK..."

For now, I recommend attempting a 1 or 2-CD rip. Just to learn to use the encoding suite. Avoid resizing above 100%. And, in the future, when size is no object, if you use an .avi player that can force an aspect ratio to 16:9 or 4:3, you may be able to eliminate the resizing from your .avs scripts. But for now, please follow the guide.

"Neutral Bicbic"

If going for a 2-CD sized rip, many folks here recommend sharp bicubic, due to DivX's tendency to soften. It's not as compressible, but, hey, you've got 4.7GB CDs. There are even sharper resizers, such as Lanczos, but that sort of stuff is for after you have some experience.

"11) Compression test @ 5.
---> b/p*f .265, 84%, .316 compress test
---> Jumping to a resolution of 704x288 changes as follows:
---> ---> b/p*f .228, 72.2%, .316 compress test. Aspect error -0.1 !
---> ---> Jumping any higher and width goes over original 720! "

As I stated ... ignore b/p*f after a compression test. You've already caught the issue about going over 100% resizing. Other than that, many recommend a second test if you resize based on the first, since the frames will have slightly different information in them at the new resolution.

"MP3 expert @ 160"

If your estimation was for 128, and your actual was 160, you will throw off the bitrate calculations used for your compression test (and b/p*f, but you're no longer looking at that value). The difference in audio filesize is 25%, which is more than you'd get using estimated/actual at the same bitrates.

"How'd I do?"

Pretty good!
jggimi is offline   Reply With Quote
Old 20th November 2002, 17:35   #7  |  Link
ScarMouth
Registered User
 
Join Date: Jul 2002
Posts: 27
Having visited a plethra of hardware stores in my area, I came to the conclusion that buying a DVD player outright would be a mistake since the most expensive wouldn't play my SVCD's and the cheapest ones somehow robbed me of quality.

So I instead I acted upon a long time ambition and built a PC for my entertainment system that was capable of playing ALL the formats.

With this in mind, it no longer became an issue to create VCD's or SVCD's since I could play from the HD or from files dropped onto a DVD.

And with the nice public library system at my disposal...Ya know, since the public owns the library, does that not mean that their DVD's also belong to us contrary to hollywood's asertion of one DVD, one owner....Anyway...:)

Knowing that the MAESTRO option was still above me, I decided to just make good streams. I've done a bit of stuff with DD2SVCD and know that spreading an image to 2 or 3 disks will provide a higher quality flick than one disk alone. However, I wanted ONE image not the 2 or 3 that were spanned across CD.

Heck I once tried to create a HUGE 1 VCD image that I planned to slap onto a DVD but the software bombed out. I suppose the image I created was over the 2gig ISO limit...

Without really digging into the various option offered on DOOM9, I stumbed across the GNOT and thought it offered a neat solution. Is there perhaps a better solution that I blindly missed?


Anyway, the guide seems to suggest the quantity of your CD's is the most important issue.

Perhaps that would also be my issue as well were it not for the DVD burner:)

However, I did see mention of BPF and that building a 2 CD image implied a good quality rip with good compression. Since I wanted a good quality rip with good compression, I tried to use that as a basis for my rip.

I saw that each of the CODEC (DIVX 3,4,5) had their own baseline BPF. Being always obsessed with the latest, I chose the DIVX5 scenario. It suggested a baseline BPF of .17 for one image and .25 for two.

So, I used the .25 as my baseline. I did want a higher audio quality and noticed that the BPF changed whenever I went to a higher quality. I believe that the BPF was in fact tied to audio quality so for now, I left quality at 128 so to be inline with the .25 spec. Later when actually encoding, I jumped to the 160 rate thinking that I already locked in a video quality. Yes?

I have done a few GNOT scenarios.
I have a Harry Potter steam that absolutely looks brilliant but, the size of the file is 2.6 gig! Nero didn't like this size too much since it exceeded the 2 gig ISO limit. Also, I was not focusing on BPF so am really not sure how I arrived at my destination.. I did though do the rip at lower BPF's and noticed a definate loss of quality. From that experience, I became more aware of BFP. But at the same time, I did not consider the "slider" and how changing the output resolution would affect the final stream. Perhaps its just me but the guides don't spend enough time discussing this issue.

I'll probably redue the Harry Potter rip with the above considered and definately break the AVI better than I did this one (so to be within ISO bounds).

I suppose I have the required selection of players for my leisure PC so I guess I can force playback at 16:9 or 4:3.

Regards Cropping...I suppose I want to AUTO crop to not encode the blank bars above and below the stream but should avoid the SMART CROP ALL because I'm running the image through a PC?

How can I benefit from this?

Can I force GNOT to do 1 pass? Beyond the time isse, do I get better quality from 2 pass? Time negated, why choose one over the other?

Um editing the scripts seems rather intimidating.

I suppose the whole idea of the GNOT is to create a good rip which it to be spanned. But it is NOT (I assume) geared towards VCDish formats since the output is AVI which must be later re-encoded by MPEG spec. Yes?

I apprecite your time (both of you) taken to help me understand this!

Scar...

Last edited by ScarMouth; 20th November 2002 at 18:02.
ScarMouth is offline   Reply With Quote
Old 20th November 2002, 18:54   #8  |  Link
jggimi
Moderator Emeritus
 
jggimi's Avatar
 
Join Date: Dec 2001
Location: The NW corner of Lake Erie
Posts: 5,552
Gordian Knot produces, as final output, SBC (DivX 3) or DivX 5 .avi files, not VCD (MPEG1) or SVCD (MPEG2) output. If your intent is to produce VCD or SVCD, going to DivX as an interim step will only reduce final output quality.

GKnot, in default mode, will calculate an average bitrate, based upon data you provide it: 1) intended file size, either by selecting CD size and CD quantity, or by typing in a size in MB or KB, 2) the size of 1 or 2 audio tracks, either estimated (by bitrate) or actual (by selecting the files), 3) audio/video interleaving overhead, 4) other files you want included.

The default mode is called "Calculate Average Bitrate." You can use the other mode of the calculator, by selecting "Calculate Avi File Size" on the bitrate tab. In that case, you provide a bitrate, and GKnot will display the total file size in KB/MB.

In either case, GKnot will use the bitrate value, recalculated based on changes between estimated and actual audio track sizes, or if you encode end-credits separately. GKnot creates an .avs script, and uses it as input to Virtual Dub, twice, for two-pass encoding. Along with other settings (such as B-Frames, GMC, which pass it is), the average bitrate requested is given to the codec.

In a 2-pass encoding, the first pass produces a frame-by-frame analysis of the video. Motion, complexity, and other information is collected in a .log file. A dummy .avi file is produced, but there is no video in it. The .log file is then used as input to the second pass, which produces the highest possible quality video using the bitrate requested as the average. Some frames will get more than others, based on the input from the first pass. DivX will sometimes produce a video with a lower average bitrate than the one requested, if it reaches maximum quality with the available bits.

The codec has two methods for 1-pass encoding: constant bitrate, which, like 2-pass, uses bitrate as the key input, and quality based, which uses a percentage or Quant. The latter is used by Gknot only for end-credits encoding. 100% quality (or Quant 2) is effectively the same as 2-pass encoding at extremely high bitrates -- the codec will use all available bits without regard to bitrate and therefore without regard to final filesize.

Back to using GKnot:

You seem absolutely stuck on b/p*f. Don't be. If, and only if you intend to produce an .avi file at a specific size, then use it, but only until you complete your first compressibility test. After that, use the %LOAD from tests, and ignore b/p*f.

If size is not the issue, then 1) Give up on b/p*f and compressibility tests, and 2) do single-pass encodings at 100% quality. It's not that difficult to run a single-pass quality based encoding using the .avs output of Gknot with Virtual Dub. You should still use GKnot to transcode the audio -- then "Save" the .avs file instead of "Save & Encode." Doom9 even has a guide: http://www.doom9.org/divx5-vdub.htm

You'll need to mux the audio with the video to create your final .avi file, using Nandub. Look for "Multiplexing AVI and MP3" in this guide: http://www.doom9.org/virtualdub_procedures.htm
jggimi is offline   Reply With Quote
Old 21st November 2002, 02:57   #9  |  Link
JohnMK
Registered User
 
Join Date: Sep 2002
Location: Seattle
Posts: 551
Quote:
Originally posted by jggimi
"Neutral Bicbic"

If going for a 2-CD sized rip, many folks here recommend sharp bicubic, due to DivX's tendency to soften. It's not as compressible, but, hey, you've got 4.7GB CDs. There are even sharper resizers, such as Lanczos, but that sort of stuff is for after you have some experience.
Your advice here is currently out of date and in fact inaccurate. One can not necessarily use a sharp resizer on a 2 CD movie. There are often times when both CDs will not be full resolution, for instance 592x256 or some such. You're better spending the bits upping the resolution one notch. Keep the resizer neutral unless and until at full resolution you're at over 65% compressibility or so.

So to sum it all up, if you want some good general advice: always use neutral bicubic, even for 2 CD encodes.
JohnMK is offline   Reply With Quote
Old 21st November 2002, 04:06   #10  |  Link
jggimi
Moderator Emeritus
 
jggimi's Avatar
 
Join Date: Dec 2001
Location: The NW corner of Lake Erie
Posts: 5,552
Well, John, I could be wrong. I often am. I'm not the self-styled "Master of Misinformation" for nothing. :-)

I know that Doom9 states he prefers Neutral Bicubic resizing, as you do. And you're right, sharp resizers affect compressibility.

There are many people who prefer SBC or XviD to DivX 5 for reasons of subjective sharpness. The development in the last couple of months of the LancsozResize AviSynth filter as a possible improvement over sharp bicubic resizing is indicative of the quest for sharper video images. Its rapid adoption from plug-in filter to built-in is testament to its popularity, at least.

Of course, if Scarmouth decides to delete resizing from his .avs scripts, then our argument is moot, anyway.
jggimi is offline   Reply With Quote
Old 21st November 2002, 04:16   #11  |  Link
ScarMouth
Registered User
 
Join Date: Jul 2002
Posts: 27
Naaa, I dont feel stuck on BPF.
I think of it simply as a point of reference.

I guess my whole issue is that the GNOT seems to have been created as a way of providing a good rip on limited space.

If I'm going to create a rip, I may as well try to acheive a good rip AND use as little space as possible.

Also 4 me, the KNOT acts as a focal point: all of the dynamics involved to make a good quality rip are there. Yet all of the adjustments are not set up as a 1-2-3, ya gotta play and because I played, I found that I could shift the focus a bit and do stuff taylored to my situation. In fact, my playing here will devinately help when I move to MAESTRO:)

From my experience, a 2 CD set gives you a good quality rip.

Why not use that compression standard my rips? I do understand that I'm not using GNOT as originally intended but then, TEFLON was an accident too:) I've made some really nice vids this way and if I ever decide to do something more serious with the rip, I can always do the rip again.

Thanks for the reference to the guides; however, the author biased it towards the 2-pass method
Quote:
2-pass: The 2 ...This is the mode to go if you want to achieve a certain size in the end so this is the mode we're going to use.
So I wonder if some of the document is also regards this focus. Anyway as best I could, I used the guide to do a sinle pass quality @ 100% to see whats up.

What seems to be working for me is to:
1) Rip
2) Make DV2 and load DV2
3) Set DIV5
4) AUTOCROP then SMARTCROP
5) Bring resolution up to original (close as possible)
6) Set MP3 settings (128 1xVBR)
7) Set BFP above .25 by increasing vid size(adjusting how big file SHOULD be:) Probably hover around .300 or so
8) Look at final vid size and adjust to taste:)
9) Compression check...not too sure about bicubic but assume sharp is best considering I'm not bound to 1 or 2 CD's.
10) Make sure percentage is above 80. If not, increase size of video!
11) Run compression check again. Jump to step 10 as needed:)
12) Encode (set to 160VBR audio)

The neatest thing I imagine is since I at least understand why I'm doing what I'm doing (however misguided it is:), it really gives me a handle on the process.

I suppose I must dig into the AVS file...eh? Of course, I can (or should be able to) see all the stuff I did in the AVS, eh?

Last edited by ScarMouth; 21st November 2002 at 05:54.
ScarMouth is offline   Reply With Quote
Old 21st November 2002, 04:51   #12  |  Link
ScarMouth
Registered User
 
Join Date: Jul 2002
Posts: 27
Quote:
Originally posted by jggimi
Of course, if Scarmouth decides to delete resizing from his .avs scripts, then our argument is moot, anyway.
Are you out of your skull man?
You want me to edit my AVS file?
Just how smart do you think I be??



I just ventured into the AVS file and gotta scream, "IS THAT IT???"
The DV2 file is a bit more intimidating; don't suppose I do anything with it eh?

So what happens if I delete the BICUBIC filtering? Quality of vid gets real soft?

Last edited by ScarMouth; 21st November 2002 at 05:45.
ScarMouth is offline   Reply With Quote
Old 21st November 2002, 17:23   #13  |  Link
jggimi
Moderator Emeritus
 
jggimi's Avatar
 
Join Date: Dec 2001
Location: The NW corner of Lake Erie
Posts: 5,552
If you delete the resize command:

1) You must be very careful with cropping values, and no longer use autocrop/smartcrop. (More on this below.)

2) You'll need a player that can force aspect ratio to 16:9 or 4:3, such as BSPlayer, since you're not using resizing to establish the display aspect ratio.

If you were to remove the cropping line as well as the resizing line, then the video stream sent to the codec -- regardless of its display aspect ratio -- would always be 720x480 for NTSC dvds. 720 is divisible by 16, and 480 is divisible by 32. These even divisions -- factors -- are called "modulos", or "mods".

The codec must be sent video with specific modulo limits -- the color space requires even pixels for one of the two axes, I can't remember which, and, I believe I've read that the DivX macroblocks are groups of 8x8 pixels. I've also been told that some video cards can't play video under mod-16.

Now, GKnot allows for very flexible cropping values -- 1 pixel at a time, if you want. However, you've probably already noticed that the default resizing mods are 32 and 16. You can change these all the way down to mod 1 if you want, but you're liable for codec failure if you select an invalid resolution. I've never gone below mod 16 or mod 8, myself. If you search the forums, you can find quite a number of people complaining that their encodings didn't work, and, once their Gknot.log files were examined, it was determined they set the mod numbers too low and picked an illegal resolution.

So, if you remove or comment out the resizing command, you'll want to be sure your crop values leave a valid resolution. You could eliminate cropping, too, but then you'd be encoding letterboxing.

Last edited by jggimi; 21st November 2002 at 17:25.
jggimi is offline   Reply With Quote
Old 23rd November 2002, 07:09   #14  |  Link
ScarMouth
Registered User
 
Join Date: Jul 2002
Posts: 27
Quote:
Originally posted by jggimi
If size is not the issue, then 1) Give up on b/p*f and compressibility tests, and 2) do single-pass encodings at 100% quality. It's not that difficult to run a single-pass quality based encoding using the .avs output of Gknot with Virtual Dub. You should still use GKnot to transcode the audio -- then "Save" the .avs file instead of "Save & Encode." Doom9 even has a guide: http://www.doom9.org/divx5-vdub.htm
I visited the guide. Whys does it suggest that PSYCHOVISUAL EFFECTS are broken?

If I've already specifed CROP/RESIZE in the AVS file, should I do it again here? If I don't?

Use the above only for VIDEO? I mean, shut down the AUDIO processing then using the same AVS, process audio with GKNOT?
ScarMouth is offline   Reply With Quote
Old 23rd November 2002, 11:34   #15  |  Link
manono
Moderator
 
Join Date: Oct 2001
Location: Hawaii
Posts: 7,406
Hi-

Why does it suggest that PSYCHOVISUAL EFFECTS are broken?

Don't know. I don't think I'd use Strong or Extreme, but as far as I know Light and Normal should be fine, if you want to use PE.

If I've already specifed CROP/RESIZE in the AVS file, should I do it again here?

No.

Use the above only for VIDEO?

Yes.

...process audio with GKNOT?

Either through GKnot, or directly, using BeSweet or HeadAC3he.
manono is offline   Reply With Quote
Old 23rd November 2002, 23:44   #16  |  Link
ScarMouth
Registered User
 
Join Date: Jul 2002
Posts: 27
1-pass quality encode

Well fellas,

I did just that.

Used GKNOT to to create my AVS and my VBR audio stream.
I used VIRTULDUB to create my VID stream. DIVX5.

Then I used NANDUB to multiplex the AVI and MP3 stream.

Seems to have worked well:)

However, the file is perhaps bigger that I wanted:(
But I can be assured that the quality is great:)

YING YANG (:) get it?? hehehe!
ScarMouth is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 14:50.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.