Avisynth: MA Script for a one-Movie DVD ? - Page 5

Dialhot · #81 09-13-2004, 04:19 AM

Quote:

Originally Posted by audioslave

Do you use the same script for both menus+extras and for the main movie? I mean, sometimes the menus and extras are 4:3 but the movie is 16:9...?

Your answer is in your question.
Let me ask it differently : do I use the same script for a 2.35:1 source and a fullscreen one ? According to you... ?

audioslave · #82 09-13-2004, 05:57 PM

@Dialhot

No, sorry for asking my dumbass questions.

But, is there an easy way of using different scripts for different parts of the DVD?

Dialhot · #83 09-14-2004, 06:44 AM

Quote:

Originally Posted by audioslave

But, is there an easy way of using different scripts for different parts of the DVD?

I guess you mean "with DVD-Rebuilder" ? No there is not.
When I use DVD-RB I do not do any crop/resize. I just put filters in the script and let DVD-RB take care of other issues.

J-Wo · #84 09-17-2004, 08:14 PM

Okay, here's an abridged summary of the scripts mentioned in this thread. Can anyone suggest which one they're most pleased with? (All filters are placed between resizing and addborders)

incredible, Feb 23, 2004

Code:

Deen("a2d",1,3,5) # Only if spatial filtering is really needed
Temporalsoften(2,3,5,20,2)
Unfilter(60,0)

dialhot, Aug 31, 2004

Code:

Undot()
Deen()

boulder, Sep 04, 2004

Code:

RemoveGrain() # note, mode=2 is default
RemoveDirt(mthreshold=100)

jorel, Sep 04, 2004

Code:

undot()
asharp(1,2)
deen()

Then on Sep 11, 2004 Black Prince said "I'm going with asharp-deen " but I don't know if he really meant undot-asharp-deen as shown above!

Also:

Quote:

Originally Posted by incredible

Jorel, try instead of deen a sample using removedirt(...) as suggested by boulder. I personally avoid using a spatial filter when treating (good) DVD sources. Maybe undot() or Removegrain at light settings but nothing else.

incredible have you got a sample script for this suggestion?

There is also an active removegrain+removedirt thread in the avisynth forum. I think this thread is more relevant for me because I too am going for 1-to-1 dvd movie encodes (although I haven't tried dvd-rebuilder yet. Better than dvdshrink I presume?). From my own experience I've had good results with removegrain as an undot replacement. But removedirt at default settings can cause "twinkling" of pixels which I find very bothersome.

Like Jorel I'm just trying to keep this thread alive!

black prince · #85 09-18-2004, 08:17 AM

@J-WO,

Quote:

Then on Sep 11, 2004 Black Prince said "I'm going with asharp-deen " but I don't know if he really meant undot-asharp-deen as shown above!

Yes, I meant undot-asharp-deen. Like you, I'm waiting until there's a
winner. With undot-asharp-deen, encoding is much faster and quality
is similiar to MA script.

I believe it takes time to test these
scripts, so that why no quick resopnse. This post is still alive, but
comparing is subjective.

-BP

J-Wo · #86 09-18-2004, 10:53 AM

Thanks for the opinion blackprince. For me I'm not entirely convinced that any script with deen is so good anymore. I think deen adds more macroblocks than it removes... I'm also going to experiment with replacing undot with removegrain to see if that makes an improvement. Hopefully with time we can all decide on a new definitive optimal script!

But just to add more fire to the flame, here's my own script that I used in the past to handle grainy dvd sources:

Code:

removegrain(mode=8) # mode=2 may actually be better/sufficient
TemporalSoften(2,4,6,6,2)
DCTFilter(1,1,1,1,1,1,0.5,0)

J-Wo · #87 09-18-2004, 02:51 PM

okay guys, I did some more testing of my script vs jorel's script. I have to say that for me personally, I'm not pleased with any undot.deen combination. It just creates too many macroblocks and such, especially on smooth backgrounds. I also realized that DCTFilter isn't needed, as I guess everyone else has already discovered. Visually I don't see any difference with it in or out but it does increase filesize which is not a good thing. Finally, removegrain(mode=8) tends to remove too many details and also create some macroblocks, so for now I'll stick to default (mode 2). So here's my personal script, which is excellent for grainy and clean DVD sources:

Code:

removegrain(mode=2)
#asharp(1,2)                     #optional
TemporalSoften(2,4,6,6,2)

I found that asharp does inrease compression a bit but I do find it creates some minor macroblocking in the background. It's hard to notice, but I prefer to have it off because I feel it preserves more details that way.

Hope you guys can provide some comments or suggestions!

incredible · #88 09-18-2004, 03:11 PM

Quote:

Originally Posted by J-Wo

Code:

removegrain(mode=2)
#asharp(1,2)                     #optional
TemporalSoften(2,4,6,6,2)

Nice to see testing you

One thing: Asharp1,2 does "plain" to much! Check hairs and skins!
Only use 1,4 or "try" 1,3 .... as everything below smoothes non-edged areas

BECAUSE: Asharp is NOT only an adaptive sharpener (based on edgedetection) but also a smoother! Means when using 1,2 beside sharpen edges, also normal areas and for shure surfaces will be plained. And that after an already processed spatial work *wow

Try for TV purposes also Unfilter(60,0) (or was it (0,60) ... Well I check that) instead of asharp() .

J-Wo · #89 09-18-2004, 07:36 PM

thanks inc, I was confused why I wasn't seeing much sharpening with asharp but instead some lost details! Let me know which unfilter settings I should try. However I'm not too sure if I need to do any sharpening, as my source is anamorphic 16:9 and so is my output.

So I finally figured out how to implement Dialhot's suggestion (on pg 1 of this thread) which requires no lanczos or bicubic resizing, which will always loose sharpness/details. I refined his coding however so it makes a little more sense and is not PAL/NTSC specific:

Code:

mpeg2source(whatever.d2v)
crop(8,x,-8,-x)    
# your filters here
addborders(0,x,0,x)

where x=0 for 4:3 fullscreen source
x=16 for 1.85 anamorphic source
x=32 for 2.35 anamorphic source

So when I finally figured out what all these numbers mean. In the first line you are cropping 8 pixels off the left and right side which brings the resolution down to 704x480, an acceptable dvd resolution. And then values for x represent how much is cropped from the top and bottom. These values represent overscan=2 parameters, so you can feel free to inc or dec x according to what your TV can handle. Finally the addborders line adds black bars to the top and bottom (i.e. letterboxes) so again you get a final res of 704x480.

I got a pm from someone who suggested this be made a sticky in the avisynth forum. What do you guys think?

nicksteel · #90 11-21-2004, 02:02 PM

Quote:

Originally Posted by Dialhot

Quote:

Originally Posted by incredible

mpeg2Source("xxxxxxxxxxxxxxx.d2v")
Your resizing here()
Deen("a2d",1,3,5) # Only if spatial filtering is really needed
Temporalsoften(2,3,5,20,2)
Unfilter(60,0)
Addborders(xxxx)

That is quite the script I use.
As I do al my targets in anamorphic, and all my source are anaporphic, I don't have even to do any resising. I just crop the black borders and put them back at the end of the process.

The borders (top, bottom) are always the same (you can check that in DVD2AVI:

(0,0) when the movie is fullscreen
(16,16) when it is 1.85
(72,72) when it is 2.35.

And the script is :

Code:

mpeg2Source("xxxxxxxxxxxxxxx.d2v")
Crop(8,0,704,576) # for full screen
#Crop(8,16,704,544) # for 1.85
#Crop(8,72,704,432) # for 2.35

Fluxsmooth()
TemporalCleaner(5,10)
Undot()
DCTFilter(1,1,1,1,1,1,0.5,0)

AddBorders(0,0,0,0) # for fullscreen
# AddBorders(0,16,0,16) # for 1.85
# AddBorders(0,72,0,72) # for 2.35

Letterbox(16,16,16,16) # overscan=2

That's all.

Note: I only did clean sources for the moment; Badly noisy ones can be processed with a spacial like deen, as inc suggested.

Can one Telecide/Decimate the above to gain DVD space (Long movie), or is it best to use MA when CQ falls below 80? Seeking best quality.

Telecide(order=1, guide=1, gthresh=10, post=2, vthresh=50, dthresh=7, blend=false, show=false)
Decimate(cycle=5, mode=0, threshold=0, threshold2=3, quality=2, show=false)

Dano · #91 11-21-2004, 11:13 PM

This is a slow but clean script I use for 1 cd per TV episode cap:

converttoyv12()
telecide(order=1)
decimate()
GripCrop(704, 480, overscan=2, source_anamorphic=false)
GripSize(resizer="LanczosResize")
DeGrainMedian(limitY=5,limitUV=10,mode=0)
VagueDenoiser(threshold=1.0,method=3,nsteps=6,chro maT=1.0)
GripBorders()

Tell me what you think.

http://home.comcast.net/~dfsuth/images/trek.mpv

Dialhot · #92 11-22-2004, 03:53 AM

Result are good but perhaps a little too smooth. VagueDenoiser is know to be very powerfull and you should even try without DeGrainMedian.

Boulder · #93 11-22-2004, 06:56 AM

Quote:

Originally Posted by nicksteel

Can one Telecide/Decimate the above to gain DVD space (Long movie), or is it best to use MA when CQ falls below 80? Seeking best quality.

If the video is telecined, of course. In fact IVTC should be done!

IVTC and MA script do not interfere, i.e. you can do the IVTC and use the MA script. Although I wouldn't use it for DVD encodes, too much filtering IMHO.

nicksteel · #94 11-22-2004, 08:52 AM

Mpeg2Source("*.d2v")

Telecide(order=1, guide=1, gthresh=10, post=2, vthresh=50, dthresh=7, blend=false, show=false)

Decimate(cycle=5, mode=0, threshold=0, threshold2=3, quality=2, show=false)

Crop(8,60,704,35

removegrain(mode=2)
asharp(1,4) #optional
TemporalSoften(2,4,6,6,2)

AddBorders(00,60,00,62)

The actual script I'm trying (Phil's one disk DVD script modified for NTSC and my source).

The source is 2.35 (16:9 Film in DVD2AVI) with anamorphic in, anamorphic out (TMPGEnc Video Aspect Ratio set to 16:9). Length is 200 min. CQ ends up as 80.5 with CQMatic. Sound is Headac3he 192/48.

jeo · #95 11-22-2004, 09:39 AM

@ nicksteel

Quote:

Length is 200 min. CQ ends up as 80.5 with CQMatic. Sound is Headac3he 192/48.

i don't saw your media size choosed, only showing that Phil's one disk DVD script.
is for kdvd right ?

nicksteel · #96 11-22-2004, 09:48 AM

Quote:

Originally Posted by jeo

@ nicksteel

Quote:

Length is 200 min. CQ ends up as 80.5 with CQMatic. Sound is Headac3he 192/48.

i don't saw your media size choosed, only showing that Phil's one disk DVD script.
is for kdvd right ?

KDVD Full 4.70GB DVD-5

I also use DVDShrink 3.2 to fit final size (98.5 for this one).
Deep Analysis, Sharp(default)

Boulder · #97 11-22-2004, 10:39 AM

Quote:

Originally Posted by nicksteel

The source is 2.35 (16:9 Film in DVD2AVI)

If the source is really FILM, you should enable Force FILM in DVD2AVI, save the project file and remove the Telecide and Decimate lines from your script.

Your borders also look weird, where did you get the values?

nicksteel · #98 11-22-2004, 10:59 AM

Thanks, Boulder.

Will try ForceFilm. I've read other places that it was better to IVTC in avs. My borders are exactly from Clip in TMPGEnc. When I use DVD2AVI, I cannot set exactly.

Is it better to just use clip values from DVD2AVI that include some of the top/bottom borders but does not clip video area than the way I am using?

When I use DVD2AVI, I have to choose clipping a few pixels of video or leaving a few pixels of border.

Also,

I usually just use Phil's script and process interlaced as anamorphic when things fit.

Dialhot · #99 11-22-2004, 11:12 AM

Quote:

Originally Posted by nicksteel

Will try ForceFilm. I've read other places that it was better to IVTC in avs.

When a source is FILM then it is not telecined ! It is "soft pulled down" and that leans that it is in fact a progressive 23.976.
When a source is VIDEO, then it is telecined and then you have to use a IVTC in the avs.

Boulder · #**100** 11-22-2004, 11:21 AM

Why use DVD2AVI for cropping?

I open the script in VDub, find the crop parameters manually using the null transform filter for example and enter them in the script (Crop(left,top,-right,-bottom). Then I open the d2v file in FitCD, feed those crop values and it gives me the correct resizing and AddBorders parameters which I add to my script. Even as we are doing DVDs, there's still no reason not to use some kind of overscan as compressibility increases quite nicely.