digitalFAQ.com Forums [Archives]

digitalFAQ.com Forums [Archives] (http://www.digitalfaq.com/archives/)
-   Avisynth Scripting (http://www.digitalfaq.com/archives/avisynth/)
-   -   CQ vs. CQ_VBR ... VERY INTERESTING... (http://www.digitalfaq.com/archives/avisynth/1910-cq-vs-cqvbr.html)

SansGrip 01-12-2003 10:50 PM

Quote:

Originally Posted by kwag
The best I've been able to do with the GOP without sacrificing compression and eliminating "almost all" flashing is 1-24-2-1-24.

Isn't that the one we started with? :D

SansGrip 01-12-2003 10:53 PM

Quote:

Originally Posted by kwag
I think CQ with the latest changes is the best mode, even for 352x240 :D

CQ mode with the new/newer GOP and beta-1a is definitely better than CQ_VBR with the same GOPs and matrix. Question is: is it better than CQ_VBR with the original GOP and matrix? ;)

kwag 01-12-2003 11:03 PM

Quote:

Originally Posted by SansGrip
CQ mode with the new/newer GOP and beta-1a is definitely better than CQ_VBR with the same GOPs and matrix. Question is: is it better than CQ_VBR with the original GOP and matrix? ;)

Oh no, no no, please :mrgreen: Is this ever going to reach a stable level :hammer: :bugeyes:
I will test that, but for the next revision :!:
I think we have a very solid quality from 352x240 all the way to 528x480 and even 704x480 if necessary. So the question is: "Do I cut the final parameters as they are, and post all templates with the new updates :?: :?: " :wink:

-kwag

SansGrip 01-12-2003 11:15 PM

Quote:

Originally Posted by kwag
"Do I cut the final parameters as they are, and post all templates with the new updates :?: :?:"

Yep.

kwag 01-12-2003 11:19 PM

Revising parameters and templates. I'll post in the "What's happening" in about half hour, after I upload all templates. :wink:

-kwag

SansGrip 01-12-2003 11:24 PM

Quote:

Originally Posted by kwag
I'll post in the "What's happening" in about half hour, after I upload all templates. :wink:

Hey... does that mean we can go back to encoding movies? 8O :D :mrgreen:

black prince 01-12-2003 11:24 PM

@Kwag,

Kwag wrote:
Quote:

So the question is: "Do I cut the final parameters as they are, and post all templates with the new updates "
Yes!! :D There are some who are confused and want, for now, a stable
set of Templates. We need to be updated with the process from GripFit
to Tmpgenc. The rest is the same (e.g. Headac3he, bbMPEG, and
VCDEasy.). Oh!! I forgot about the SansGrip's new file predictor
(manual) method. We need to catch-up to the latest and greatest stuff.
It would be nice to know what you'll be working on in the future (e.g.
filters to target I, P, or B frames for compression). Or, a return to
solving Gibbs, or maybe a Guide forum or .... the list just grows. :D

-black prince

SansGrip 01-12-2003 11:32 PM

Quote:

Originally Posted by black prince
There are some who are confused and want, for now, a stable set of Templates.

Some? I think we're all confused. I know I am :? ;).

Quote:

Oh!! I forgot about the SansGrip's new file predictor (manual) method.
As far as that goes, I think it's a good way of getting the most accurate prediction possible given the current state of affairs. However, I'm not sure it's best for manual use. I designed it mainly for inclusion in the next KVCDP.

Quote:

It would be nice to know what you'll be working on in the future (e.g. filters to target I, P, or B frames for compression). Or, a return to
solving Gibbs, or maybe a Guide forum or .... the list just grows. :D
Tell me about it. You should see my to-do list 8O ;).

syk2c11 01-12-2003 11:54 PM

For the LATEST template (beta-1a, 1-24-2-1-24), do we need to check "detect scene change"??

kwag 01-13-2003 12:04 AM

Quote:

Originally Posted by syk2c11
For the LATEST template (beta-1a, 1-24-2-1-24), do we need to check "detect scene change"??

No :!: . It should be "unchecked" on all templates.

-kwag

kwag 01-13-2003 12:26 AM

Templates updated !
 
Done :ole: :ole:
http://www.kvcd.net/forum/viewtopic.php?t=2305

Let me know if I screwed up somewhere, because sometimes :imstupid:

kwag 01-13-2003 12:27 AM

Quote:

Originally Posted by SansGrip
Quote:

Originally Posted by kwag
I'll post in the "What's happening" in about half hour, after I upload all templates. :wink:

Hey... does that mean we can go back to encoding movies? 8O :D :mrgreen:

YEEEEAAH :yippie:

kwag 01-13-2003 12:32 AM

Quote:

Originally Posted by black prince
@Kwag,

Kwag wrote:
Quote:

So the question is: "Do I cut the final parameters as they are, and post all templates with the new updates "
Yes!! :D There are some who are confused and want, for now, a stable
set of Templates. We need to be updated with the process from GripFit
to Tmpgenc. The rest is the same (e.g. Headac3he, bbMPEG, and
VCDEasy.). Oh!! I forgot about the SansGrip's new file predictor
(manual) method. We need to catch-up to the latest and greatest stuff.
It would be nice to know what you'll be working on in the future (e.g.
filters to target I, P, or B frames for compression). Or, a return to
solving Gibbs, or maybe a Guide forum or .... the list just grows. :D

-black prince

I think the next step is some basic organized documentation ( To SansGrip: I feel the same way you do about documenting :x ) so people have an idea of the complete process. Now that SansGrip has released GripFit, it makes it REALLY easier to encode once a bare .avs is created.

-kwag

jamesp 01-13-2003 03:37 AM

Re: Templates updated !
 
Quote:

Originally Posted by kwag
Done :ole: :ole:
http://www.kvcd.net/forum/viewtopic.php?t=2305

Let me know if I screwed up somewhere, because sometimes :imstupid:

Kwag,

Will the DVD templates and the 528x576 Mpeg2 template be updated too? Or are these not suitable for the latest enhancements.

Jim

GFR 01-13-2003 04:57 AM

Quote:

Originally Posted by black prince
Then Gibbs is a result of Tmpgenc motion search? That makes sense.
So the only way to really improve and/or remove Gibbs is to correct
Tmpgenc encoder. Hey Frankencoder, where are you!!!!! :)

-black prince

AFAIK, Gibbs is the result of quantization in the high frequencies.

Jellygoose 01-13-2003 05:53 AM

:D Ok, ok guys...

I guess we're all VERY satisfied with the results of the last month... I guess kwag should close this and the other threads that created this great brainstorming progress in a couple of days! :wink:
As for me, I already saved all the pages (38 wow!!) and love to read through them, to see you all having excellent ideas, and of course testing, testing, testing...

As for me I think all this was definetely worth the effort...

THE NEW KVCD v2 ROCKS!!!!

guys I really can't imagine where we will be in one year!!
This forum has developed to my favorite site, and I still think it's amazing and impressive that people from so many different countries share their knowledge and experiences in video encoding...
that's kinda like Captain Planet, where everybody gives a piece of power and in the end there's a SUPERHERO!!

gonzopdx 01-13-2003 06:06 AM

8O 8O 8O

GOP: 1-24-2-1-24
Matrix: KVCD Notch BETA-1a
Source: 480x480 SVCD
Encoded To: 480x480 MPEG1, 128k Sound (2x80-min)
Script:
Code:

(...plugins...)

(...source...)

LegalClip()
Crop(0,102,0,-108)
BicubicResize(448,252,0,0.6)
FluxSmooth()
BlockBuster(method="noise", variance=.3, seed=1)
LegalClip()

CLIP: Here
8O 8O 8O

kwag 01-13-2003 08:00 AM

Re: Templates updated !
 
Quote:

Originally Posted by jamesp
Will the DVD templates and the 528x576 Mpeg2 template be updated too? Or are these not suitable for the latest enhancements.

Jim

Hi Jim,

The new BETA-1a Notch matrix applies perfectly well to the KDVD templates. I only tested the KDVD Half D-1 with the new matrix, and I liked the results. I'll do more tests today, and if everything works ok, I'll update those too. The same applies to the SKVCD and the KVCDx3 (MPEG-2). They should also work better with the new matrix. If you want to do some tests on the DVD templates, just change the number of Bs from 3 to 2 and make the matrix changes.

-kwag

kwag 01-13-2003 08:17 AM

Quote:

Originally Posted by Jellygoose
:D Ok, ok guys...

I guess we're all VERY satisfied with the results of the last month... I guess kwag should close this and the other threads that created this great brainstorming progress in a couple of days! :wink:
As for me, I already saved all the pages (38 wow!!) and love to read through them, to see you all having excellent ideas, and of course testing, testing, testing...

As for me I think all this was definetely worth the effort...

THE NEW KVCD v2 ROCKS!!!!

guys I really can't imagine where we will be in one year!!
This forum has developed to my favorite site, and I still think it's amazing and impressive that people from so many different countries share their knowledge and experiences in video encoding...
that's kinda like Captain Planet, where everybody gives a piece of power and in the end there's a SUPERHERO!!

Hey thanks Jellygoose :D,

I forgot to make a note on "Version 2" on the download page. I'll do that in a while :wink:
Let's see, in a year, probably taking all the techniques we've learned here and fitting 10+ hours of perfect DVD quality on a DVD(+-)R :mrgreen:
To me, KVCD.Net is like a candy store. I chew on it all day :mrgreen:

-kwag

kwag 01-13-2003 08:33 AM

Quote:

Originally Posted by gonzopdx

CLIP: Here
8O 8O 8O

Hi gonzopdx,

Did you use prediction on that sample :?: Because I find it a little on the expensive side for 6.8MB and only 35 seconds. It should be less than ~3MB for that time :!:
Or was this made to target 3 CD-Rs :?:

-kwag

kwag 01-13-2003 12:22 PM

KDVD first V2 sample
 
Here's the first KDVD Version 2.0 Full D-1 sample. I predicted the size to fit three movies on one DVD(+-)R (Each video stream ~1.3GB ). So this is what the movies will look like, all on a single DVD-R. That's an average of ~6 hours for each DVD :mrgreen:
http://www.kvcd.net/kdvd-v2-sample.m2v

That's using the BETA-1a "Notch" matrix, and setting spoilage to 0:0, as in all other templates, and SansGrip's Blockbuster "Noise" and GripFit with all bells and whistles :mrgreen:

Script used:

Code:

LoadPlugin("C:\encoding\MPEG2DEC.dll")
LoadPlugin("C:\encoding\fluxsmooth.dll")
LoadPlugin("C:\encoding\GripFit_preview.dll")
LoadPlugin("C:\encoding\blockbuster.dll")
LoadPlugin("C:\encoding\legalclip.dll")
LoadPlugin("C:\encoding\sampler.dll")
LoadPlugin("C:\encoding\vobsub.dll")

Mpeg2Source("K:\COUNT_OF_MONTE_CRISTO\VIDEO_TS\count.d2v")
LegalClip()
vobsub("K:\COUNT_OF_MONTE_CRISTO\VIDEO_TS\VTS_01_0")
GripFit( source_anamorphic = true, dest_anamorphic=false, width=720, height=480, overscan=2)
FluxSmooth()
Blockbuster(method="noise", variance=.3, seed=1)
LegalClip()

Sampler(length=24)
## MPEG size = ((Total frames/MovieTimeInMinutes)/24) * MPEG sample file size ##

With this, all I can say for my personal taste and convenience, is "Bye Bye DivX" when DVD media becomes more affordable :wink:

Edit: Sample re-uploaded. I had mistakenly encoded using "Normal" motion estimation, instead of "High quality" :wink:

-kwag

Jellygoose 01-13-2003 12:27 PM

@gonzopdx

8O where did you get those vobs from? isn't that LOTR 2 ? Is that available on DVD somewhere already??

SansGrip 01-13-2003 01:38 PM

Prediction out by miles
 
I just finished encoding Death To Smoochy (new 352x480 template) at CQ 58.5. Target size was ~689mb, final size was ~648mb. I was able to bump the audio up from 112 to 160 8O.

kwag 01-13-2003 01:40 PM

Re: Prediction out by miles
 
Quote:

Originally Posted by SansGrip
I just finished encoding Death To Smoochy (new 352x480 template) at CQ 58.5. Target size was ~689mb, final size was ~648mb. I was able to bump the audio up from 112 to 160 8O.

How does it look :?:
That's ~6% off right?

Edit: Was it using Predictor or manual calculation :?:

-kwag

SansGrip 01-13-2003 01:44 PM

Re: Prediction out by miles
 
Quote:

Originally Posted by kwag
How does it look :?:

It looks pretty good -- this is a hard movie to compress, lots of bright colours etc. and a 16:9 ratio.

Quote:

That's ~6% off right?
Yep, 6.3% or thereabouts. Too much. I think I'll reencode at 689 / 648 * 58.5 = 62.2 and see what I get...

SansGrip 01-13-2003 01:46 PM

Re: Prediction out by miles
 
Quote:

Originally Posted by kwag
Edit: Was it using Predictor or manual calculation :?:

I used my own method, which was ~0.1% accurate before this encode :?.

kwag 01-13-2003 01:54 PM

I'm going for a 528x480 full encode of "Count of Monte Cristo" in a little while. I'm doing this one with subs, to make prediction more complicated :wink:. I just finished re-encoding the KDVD sample above, because I had set motion estimation to "Normal" :x , so I was able to increase the CQ a little. So I'm uploading the KDVD sample again, and then encoding at 528x480 with KVCDx3. I'll post an x3 sample when I get my predicted CQ value in a little while.

-kwag

gonzopdx 01-13-2003 02:18 PM

Quote:

Originally Posted by kwag
Did you use prediction on that sample :?: Because I find it a little on the expensive side for 6.8MB and only 35 seconds. It should be less than ~3MB for that time :!:
Or was this made to target 3 CD-Rs :?:

I agree, 6 megs for 30 seconds was pretty big.. I did use prediction for 2 CD's -- fits PERFECT! 1h25m~ish on each CD, 128k audio. Maybe it's so big since it's a very high action scene? It looks GREAT though! :D

Quote:

Originally Posted by Jellygoose
8O where did you get those vobs from? isn't that LOTR 2 ? Is that available on DVD somewhere already??

Yes, it's LOTR2. It's from a DVD Screener -- the 4-disc SVCD has been floating around the net for the past week or so..

kwag 01-13-2003 03:19 PM

Here's the same sample I posted above, but done with KVCDx3 targeted for one CD-R: http://www.kvcd.net/kvcdx3-v2-sample.mpg
This is a better candidate for either 352x480, or 2 CDs at 528x480 :wink:

-kwag

SansGrip 01-13-2003 03:24 PM

Quote:

Originally Posted by kwag
This is a better candidate for either 352x480, or 2 CDs at 528x480 :wink:

That's what I found with Death To Smoochy -- even at 352x480 it's not ideal quality. Strange, really, considering it's less than 2 hours long :?.

kwag 01-13-2003 03:39 PM

Quote:

Originally Posted by SansGrip
Quote:

Originally Posted by kwag
This is a better candidate for either 352x480, or 2 CDs at 528x480 :wink:

That's what I found with Death To Smoochy -- even at 352x480 it's not ideal quality. Strange, really, considering it's less than 2 hours long :?.

I guess some movies have some weird luma levels that chew up compression :!: But if it's a dark movie, where are we loosing compression :?: High spike noises on scene changes :idea:
AHA :!: , there you go SansGrip, time for a new analysis :mrgreen:
A "Scene Transition Vacuum Cleaner" .dll :bugeyes: or properly, a "transition smoother " 8)

-kwag

SansGrip 01-13-2003 03:51 PM

Quote:

Originally Posted by kwag
But if it's a dark movie, where are we loosing compression :?:

I'm not sure, but I notice that a lot of very dark movies are hard to compress (e.g. Panic Room, Resident Evil).

Quote:

A "Scene Transition Vacuum Cleaner" .dll :bugeyes:
8O Ah... I'm going to encode some movies first :P ;).

Paul0889 01-13-2003 04:12 PM

Kwag,
In the new KVCDx3 template download, shouldn't the new GOP be
1-24-2-1-24? In the new template, it's 1-5823-2-1-24

kwag 01-13-2003 04:47 PM

Quote:

Originally Posted by Paul0889
Kwag,
In the new KVCDx3 template download, shouldn't the new GOP be
1-24-2-1-24? In the new template, it's 1-5823-2-1-24

Q-29 here: http://www.kvcd.net/forum/viewtopic.php?p=326 :wink:

Paul0889 01-13-2003 05:05 PM

oh. thanks!

Paul0889 01-13-2003 05:06 PM

oh. thanks! By the way, finished Resident Evil this morning, looks GREAT!!!! A lot better than before with the old kvcdx2 template. Very impressed with the new templates.

kwag 01-13-2003 05:21 PM

Quote:

Originally Posted by Paul0889
oh. thanks! By the way, finished Resident Evil this morning, looks GREAT!!!! A lot better than before with the old kvcdx2 template. Very impressed with the new templates.

:ole:

SansGrip 01-13-2003 07:11 PM

Re: Prediction out by miles
 
Quote:

Originally Posted by SansGrip
I think I'll reencode at 689 / 648 * 58.5 = 62.2 and see what I get...

I got 650mb, up from 648mb. That's a 2mb gain from a 4-point CQ increase, over the entire movie. Something's not right...

kwag 01-13-2003 07:54 PM

My machine crashed :x
M$crap explorer did something weird, and made my machine perform like a 1.7Mhz Atari 800 8O
That's what I get for leaving XP running a week without a restart :x
I guess MsBlows Uptimes are still way WAY lower than any U*ix variant :roll:
Back to encode again :!:

-kwag

kwag 01-14-2003 12:20 AM

Look at this :)
 
Hey guys, do you remember the original "The Green Mile" samples done with CQ_VBR and the original KVCD matrix :?:
Well look again at the quality for 180 minutes on one CD-R with the new "Notch" matrix and CQ: http://www.kvcd.net/mile-v2-sample.mpg
:mrgreen:

-kwag

Edit: I just compared the sample above to a sample encoded with the "Standard" TMPEG VCD template, and there's barely any difference between the two 8O. I wonder if the "LBR" definition should really be ~200 minutes, because at 180, it looks just like a VCD :roll:
Any comments on this :?:


All times are GMT -5. The time now is 02:03 PM  —  vBulletin © Jelsoft Enterprises Ltd

Site design, images and content © 2002-2026 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2026 Jelsoft Enterprises Ltd.