digitalFAQ.com Forums [Archives] - Optimal or Girv for Divx encodes?

digitalFAQ.com Forums [Archives] (http://www.digitalfaq.com/archives/)

- Video Encoding and Conversion (http://www.digitalfaq.com/archives/encode/)

- - Optimal or Girv for Divx encodes? (http://www.digitalfaq.com/archives/encode/5205-optimal-girv-divx.html)

Reno	08-25-2003 07:22 PM

Optimal or Girv for Divx encodes?

The subject pretty much sums it up. Which one do you use, and why??

incognito

08-25-2003 07:45 PM

I just so happen to be using this one.

Code:

###################################################################### 

# Based on Optimized Script to remove DCT Blocks by Girv and DialHot # 

###################################################################### 

############## 

# Parameters # 

############## 

#Use 720 or 352 to KDVD, 528 352 to K(S)VCD 

m_width=704

#subtitle="name_of_the_subtitle.smi" # Add a # in the beginning of this line if you 

#don't need subtitles.movie="name of the movie.avi" 

myoverscan=1 # Test values from 0 to 3 here. Bigger files produce more quality but #a value too big will cause undesirable black borders on your TV. 







########## 

# script # 

########## 

AviSource("G:\Ripps\Current\anymovie.avi", false)



#Post-Process 

Blockbuster(method="noise",detail_min=1,detail_max=8,variance=0.3,seed=5823) 

ATC(1,2,4,0.5,false) 

TemporalSoften(2,7,7,3,2) 

Blockbuster(method="noise",detail_min=1,detail_max=10,variance=0.5,seed=5823) 



#Resize, overscan and borders 



gripCrop(m_width, 480, overscan=myoverscan,source_anamorphic=false,dest_anamorphic=false) 

GripSize(resizer="lanczosresize") 

GripBorders() 



#Subtitles # 

############ 

#TextSub(subtitle) # Add a # in the begining of this line if you don't need subtitles. 



converttoyuy2()

And damn it looks good.

Reno	08-25-2003 09:08 PM

Looks good. I'll have to try that soon. Re: ATC, is that another avisynth filter or is it already plugged in? Also, will I need any new filters other than those that come with the latest Moviestacker?

incognito

08-25-2003 10:12 PM

You need AVISynth 2.5 along with Blockbuster for 2.5 and ATC. The full guide is here:
http://www.kvcd.net/forum/viewtopic.php?t=5142&start=0
And the filters are here:
http://www.avisynth.org/~warpenterprises/

Enjoy. I know I did. :wink:

Dialhot

08-26-2003 03:15 AM

Quote:

Originally Posted by incognito

I just so happen to be using this one [..] And damn it looks good.

Just two things I did based more on thinking that really on experiment :

1/ do the last blockbuster after the resizing as else this one will crunch some of the noise added by blockbuster and the effect will be less efficient.

2/ the seed used for both blockbuster lines must be different else you have two time the noise on exactly the same pixels off the picture. That's a derivate of random number generation based on a seed.

vmesquita

08-26-2003 09:19 AM

Dialhot,

Great ideas! What value can I use in the second blockbuster as seed? I was thinking about trying BlindPP() before all the filters... Just some extra thoughts...

[]'s
Vmesquita

Dialhot

08-26-2003 09:32 AM

The seed is only a starting point for the randomizer, you can put whatever you want. I use to put "5823" and "5623".

incognito

08-26-2003 09:08 PM

I'm not a pro at writing scripts yet, so it should look like this?

Code:

###################################################################### 

# Based on Optimized Script to remove DCT Blocks by Girv and DialHot # 

###################################################################### 

############## 

# Parameters # 

############## 

#Use 720 or 352 to KDVD, 528 352 to K(S)VCD 

m_width=480

#subtitle="name_of_the_subtitle.smi" # Add a # in the beginning of this line if you 

#don't need subtitles.movie="name of the movie.avi" 

myoverscan=1 # Test values from 0 to 3 here. Bigger files produce more quality but #a value too big will cause undesirable black borders on your TV. 







########## 

# script # 

########## 

AviSource("G:\Ripps\animation.avi", false)



#Post-Process 

Blockbuster(method="noise",detail_min=1,detail_max=8,variance=0.3,seed=5823) 

ATC(1,2,4,0.5,false) 

TemporalSoften(2,7,7,3,2) 



#Resize, overscan and borders 



gripCrop(m_width, 480, overscan=myoverscan,source_anamorphic=false,dest_anamorphic=false) 

GripSize(resizer="lanczosresize") 

Blockbuster(method="noise",detail_min=1,detail_max=10,variance=0.5,seed=5623) 

GripBorders() 



#Subtitles # 

############ 

#TextSub(subtitle) # Add a # in the begining of this line if you don't need subtitles. 



converttoyuy2()

Dialhot

08-27-2003 03:46 AM

I think it's ok. You should try without the last convertToYuY2 as I never use it and everything is ok on my PC. It depends on the codec pack you have I guess.

vmesquita

08-27-2003 08:31 AM

DialHot,

Converttoyuy2() is needed for CCE unless you have a yu12 decoder, I guess there is one that comes with FFVFW.

[]'s
Vmesquita

Dialhot

08-27-2003 08:44 AM

yu12 or yv12 ?

Because everything is already in yv12 in your script, no ?

vmesquita

08-27-2003 09:07 AM

DialHot,

Spell error, I meant yv12. Everything in the script is already yv12, but CCE only accepts yuy2 input unless you have a yv12 decoder.

[]'s
Vmesquita

incognito

08-27-2003 10:19 PM

So if I'm using TMPGEnc I don't need

Code:

converttoyuy2()

Cool. Thanks. :D

boeddha

08-28-2003 03:19 AM

I use Girv/Dialhot script and I like the quality very much but there's not so much compression. Has anyone some ideas which filter I can ad to get a smaller filesize without affecting quality. Encoding time doesn't matter.

Thanx

Dialhot

08-28-2003 05:53 AM

Quote:

Originally Posted by boeddha

I don't know what you call "not so much compression" but having a MPEG1 KVCD at the end that is the same size than the original MPEG4 avi file, I call that "a godam' good compression" !
With old SVCD we need twice the size in MPEG2 than in MPEG4.

Btw, If you want to drop a little the size, you can change Lanczos resizer to Bicubic (and perhaps then, add a "asharp(1,4)" just after the resizing).

boeddha

08-28-2003 06:22 AM

@ Dialhot

Of course there's "godam good compression" as you say it. What I meant was in comparison with the ma script the filesize (with constant cq) is bigger. So I have to down the cq with something like 4 points in comparison withe the ma script. Maybe this wasn't so clear because my english is not so good. I didn't mean to say that kvcd doesn't give good compression. I asked if there maybe were some filters that could give me some more compression without affecting quality.

You did give me some advise so I will look If change to bicubic will give me some smaller filesize.

thanx

vmesquita

08-28-2003 07:40 AM

boeddha,

The main purpose of the Girv script is kill DCT blocks that are very common in DivX source. Since that is done in part by adding noise, it's hard to getter a better compression than the MA script. But of course we don't want better compression at all costs, we want better compression but also the better picture. I wouldn't mind lowering the CQ if I know that in the end my movie will look better because some of the weekness of the sources has (almost) gone away.
As Dialhot said, you can try bicubic, a less sharp picture will of course be more compressible...

[]'s
Vmesquita

Dialhot

08-28-2003 07:46 AM

You can also do a very slight blur on the luma (slight because MPEG4 already soften a lot).

Try : MergeLuma(Blur(0.1)) after all the other filters (just before resizing).

Reno	08-30-2003 04:44 PM

Hey guys, are you using CQ_VBR or CQ for your encodes? According to Kwag, CQ_VBR is better with Blockbuster...

Dialhot

08-30-2003 05:09 PM

I know what said Kwag about this, but blockbuster is efficient even in CQ mode. Perhaps less efficient, but efficient enought.

So, as already told in an other thread what I use is : CQ_VBR for VCD res and CQ for res above that.

kwag	08-30-2003 05:15 PM

Quote:

Originally Posted by Reno

According to Kwag, CQ_VBR is better with Blockbuster...

I said that, because that's what the developer of Blockbuster (SansGrip) said :)
He even proved it in the "CQ vs. CQ_VBR" thread.
At 352x240, in CQ_VBR, it works very well. Above that resolution and using CQ, no matter how strong you apply Blockbuster noise, it barely diminishes the DCT blocks. He provided screenshots in the thread mentioned above.

-kwag

Dialhot

08-30-2003 05:22 PM

The problem here is different : we use blockbuster for eliminate already existing blocks that are in divx. Not for avoiding new ones.

For this purpose (I can't provide screenshots, sorry ;-)), blockbuster works either in CQ than CQ_VBR mode.
Now, I agree with you that CQ_VBR is better only at VCD res, and CQ for res above.

kwag	08-30-2003 06:36 PM

Ok, I see. Then you're "spraying" the DCT blocks with noise :mrgreen:

-kwag

Dialhot

08-31-2003 05:49 AM

Yes it is :-D

Wolfi

08-31-2003 01:07 PM

Does this script look okej :arrow:

Code:

LoadPlugin("C:\Filters\MPEG2Dec3.dll") 

LoadPlugin("C:\Filters\atc.dll") 

LoadPlugin("C:\Filters\Blockbuster.dll") 

LoadPlugin("C:\Filters\VSFilter.dll") 

LoadPlugin("C:\Filters\dctfilter.dll")



AVISource("C:\movie.avi",false) 



Blockbuster(method="noise",detail_min=1,detail_max=8,variance=0.3,seed=5823) 

ATC(1,2,4,0.5,false) 

TemporalSoften(2,7,7,3,2) 

Blockbuster(method="noise",detail_min=1,detail_max=10,variance=0.5,seed=5823) 

DCTFilter(1,1,1,1,1,1,0.5,0)



gripCrop(540, 480, overscan=2,source_anamorphic=false,dest_anamorphic=false) 



GripSize(resizer="lanczosresize") 

GripBorders() 



TextSub("C:\movie.srt") 

converttoyuy2()

Is there something who should be changed :?: :roll:

//Wolfi

Dialhot

08-31-2003 02:25 PM

Quote:

Originally Posted by Wolfi

Does this script look okej

Yes it does.

bigggt

09-02-2003 08:41 PM

Hi,nice results with the divx,xvid script but what is everyone using for prediction.

I did a manual prediciton with the last movie and it was pretty good(tiny bit too big)

So whats everyone using

Thanx

kwag	09-02-2003 08:48 PM

Quote:

Originally Posted by bigggt

Hi,nice results with the divx,xvid script but what is everyone using for prediction.

Try the latest CQMatic. It should do a descent job on AVI prediction too.
Previous versions didn't work very well, because of the sampling that was used. That was changed in version 1.2.00

-kwag

bigggt

09-02-2003 08:59 PM

thanx Kwag ,i'll give it a try

incognito

09-03-2003 12:30 AM

CQMatic is doing a thumbs up job on avi encodes.

ChiefOBrei

09-05-2003 12:33 AM

I have tried this script and the results are promising. But as far as I know we have to use a different setup in TMPGEnc due to the fact that we do not have a MA script. In my opinion it makes sense to define which options should be used with this script.

- Min bitrate: 300 and Max bitrate: 2500 or should we use 0,57 avg bitrate and 2000

- Motion search precision: High quality or should we use motion estimate search

Maybe there are more options which are interesting.

Is it possible to have a sticky post with a optimal Divx script?

Thanks!

ChiefOBrei :lol:

J-Wo	09-07-2003 01:06 AM

Just to add to the last post, I'm getting funky predictions with cqmatic and tok. what are you guys using for min/max bitrate? Detect scene change? And you use calcumatic to get avg bitrate? Just wondering how you're getting accurate avi predictions.

Oh and I second the motion for a divx optimal script sticky. Perhaps dialhot could moderate it? 8)

bigggt

09-07-2003 09:53 AM

HI J-Wo

I have only done 1 movie so far but this is what i got with the kvcd 352X240 template

DEtect scene change not selected and calcumatic to get avg bit rate

Manual prediction min 300 max 1800 -CQ 80.75(file tiny bit too big)just cut out 1 min of the ending credits
TOK CQ80.693 min 300 max 1800
CQ matic min300 max 1800 -CQ 81.8 50 50]mb too big
CQ matic avg *57 min and 1800 max CQ 80.70(didn't encode but would have been the closet)

Obviously its too early to be sure but this is from my first try.

J-Wo	09-07-2003 10:31 AM

thanks bigggt. Just wondering though, do you think visual quality will be affected much by having such a low max bitrate? I thought kwag said we could lower it to 2000 because we were using the MA script, but now we're not anymore... :?: :?

bigggt

09-07-2003 12:18 PM

I think with the higher templates you would still use 2500 max(correct me if i'm wrong) but i used 1800 because it was the 352X240 template which the max should be 1800