Hey Kwag (and others!):

In both of your latest optimal scripts, you include the following line:I think I have just an example of one of those extreme cases! The movie is Tigerland by Joel Schumacher, and the entire film is shot in a very grainy handheld cam style picture. Using your optimal script for DVD and other clean sources, I get a CQ of 46.684 (using ToK). If I add SpaceDust(), the CQ goes up to 51.590 and if I add Convolution3D(preset="movieLQ") on top of that CQ becomes 56.561. So here are my questions:

- does a higher CQ necessarily mean better quality encode? To my eye the more filters I add, the softer the resulting image looks. All encodes have a lot of blockiness in the background parts of some scenes, or in dark scenes in particular. But in the encode without the extra filters the noise is more "grainy" and "sharp", and as has been previously mentioned this noise looks more realistic. With SpaceDust and Convolution3D this noise is smoothed out a little at the cost of a bit of sharpness.

- what parameters would you use for Convolution3d, STDMedian, and DCTMedianFilter? Do you recommend using all three filters at one time? If not what combo should I use?

Thanks!

Hi J-Wo,

How about trying PixieDust() and TemporalSmoother(1,2), together with the merge filters only. That should do a very good job, without blurring the picture too much.
Something like:


-kwag

Hey guys,after you put the script through tok and get your CQ,when you load the avs into tmpge do you still need to load the kvcd template also or just leave as is.

Thanx

This may be a dumb question. But how can you use these filters with moviestacker. Do you have to enter them manually ?

Not a dumb question at all :D.
Yes, you have to enter them manually, until muaddib adds the filters to the next release of MovieStacker :wink:

-kwag

There are no templates anywhere on ToK.You set your parameters ( MIN, MAX, etc. ) and ToK makes a KVCD automatically :D
The notion of a "template" doesn't exist on ToK :wink:

-kwag

where do you get these filters.

OK, so i also have my latest optimal script :lol: here it goesGripCrop(480, 480, overscan=2, source_anamorphic=false)
GripSize(resizer="BicubicResize")

unfilter(70,70)
Blur(0.2)
SpatialSoften(0,0,2)
MergeChroma(blur(1.58))
MergeLuma(blur(0.2))
GripBorders()
LegalClip()


Blur and sapatial are within Avisynth you make your owen opinion about it play with the number as you like. unfilter(70,70) my be too much for some of you. This worked for me :wink:

Just a small request, but could you add a list of filters to your sticky post? Maybe even link to the versions that you are using so we can download them easily. I think most of the "regulars" here know where and what most of the filters are, but thats one of the main things to figure out as a newbie, and then the whole version skew problem that happens from time to time.

-BC

Hi bytecode,

Did you check the "Software Downloads" link on the main page at www.kvcd.net :wink:

-kwag

Sorry Kwag , when tok is completely done and you get your CQ and then start tmpge to do the final encode i was wondering do you just load the avs or do you need to load a template also like in Red-M's guide.

I know its a stupid question but when your just learning and are computer stupid like myself once you add any kind of extra step it sort of throws things off a little. :?

your guys help is always greatly appreciated

thanx

Tom

Hi Tom,

No, when ToK starts to encode, it starts to encode :D. That's it! You don't have to do anything until the process is done. Then you mux/burn and go watch TV :D

-kwag

you can compare ToK with DVD2SVCD, except, that you do the audio source yourself. just set the bitrate and ToK does all the work, that is nessessary to get your kvcd

Hi Kwag,i am trying not to be annoying :D but i'm still not sure,i know tok does everything but all i do with it is to get the CQ(no encoding only prediction) because it takes too long to do the whole encode)

When you shut down tok and open tmpge to do the whole movie do you open the video avs that tok made or do you put the avs script that you originally loaded into Tok and then load a the template.

The reason i ask is beacuse i checked and the only difference is in the GOP tab under the part where it says Number of P picture in GOP.

THe tok avs is 24
and your template when loaded from Tmpge is 5823

Will this make a difference

By following red m's guide it says after you do the file prediction you open tmpge and then load close to the same template you used when you made your script with moviestacker,i was just wondering is this necesaryafter using tok.

Like i said the only difference is in the GOP tab

Ifd this don't make any sense(because i have trouble reading it myself)
just tell me to shut-up :D and try them both.

Thanx again

Should I be able to fit one movie on one CD with this latest script? For example "15 Minutes" which is about 114 minutes long... I'm currently using a resolution of 528 x 576.
What filter am I supposed to use for deinterlacing? Do I use DecombLegacy or Bob?

Hi audioslave,

Yes, you should probably be able to fit an ~2 hour movie onto 1 disc using the latest script and the x3 template, depending on your source material (personally, I've even found a CQ level of as low as ~50 to be acceptable quality on a standard TV using the latest optimal script...my tastes may differ from yours though). For deinterlacing, you can either use Force Film in DVD2AVI (if your source is = or > than 95% Film), or you can use DecombLegacy in your avisynth script. For 23.976fps material, use the following lines right after your Mpeg2Source line:

happy encoding,
-d&c

@ dazed&confused

Thanks for the info on the deinterlacing issue! I guess I'll have to put the movie (15 Minutes) on 2 CD's since the CQ value for 1 CD got a low as 42! 8O I'd rather prefer quality over a 1 CD movie. But that's just me :wink: ...

Actually, if you want strict Deinterlacing, you'd want to use
instead of

@ smoochie3

Yes, certainly. I guess I forgot to mention that I'm currently encoding PAL movies only. :? For the moment, that is. So, I'm using the FieldDeinterlace() command. Thanks anyway :)

What is the difference between FieldDeinterlace() and Telecide()/Decimate()?

FieldDeinterlace() will keep the framerate you start with (29.97, or 25), Telecide()/Decimate() will Inverse Telecine (IVTC) your starting framerate to 23.97 and should only be used if you know that the source was originally 23.97 otherwise you will have jerky playback.

Where can i find the filter STMedianFilter?

Sorry if this question has been asked before but this tread is getting pretty big and i din't think it was a good idea reading every one to find it.

Thanks

hi p00pz,welcome. :)

http://www.trbarry.com/STMedianFilter.zip

:!:

Thanks for the quick relpy and for the welcome.
Another thing do you know were I can find DCTFilter? Or would it just be easier to if you pointed me to a thread or site that had a list of all the filters and links to download them?

Thanks alot! :)

ok,.....

http://www.kvcd.net/forum/viewtopic.php?t=3647


:)

with this script in the newest version of tok it takes 15 minutes for the full sample

LegalClip()
GripCrop(352, 480, overscan=1, source_anamorphic=false)
GripSize(resizer="BicubicResize")
# SpaceDust() # Optional - for some "not so clean" DVDs.
unfilter(50,50)
TemporalSmoother(1,2)
mergechroma(blur(1.58))
mergeluma(blur(0.2))
# Convolution3d or STDMedian or DCTMedianFilter for extreme worse cases.
GripBorders()
LegalClip()

With the newest script
LegalClip()
GripCrop(352, 480, overscan=1, source_anamorphic=false)
GripSize(resizer="BicubicResize")
STMedianFilter(10, 50, 0, 0, 10, 50)
# SpaceDust() # Optional - for some "not so clean" DVDs.
unfilter(50,50)
temporalsmoother(1,2)
mergechroma(blur(1.58))
mergeluma(blur(0.3))
# Convolution3d or DCTFilter for extreme worse cases.
GripBorders()
LegalClip()
it takes 2 hours to do just the full sample in tok

Does this seem right

thanx

hi bigggt :)

you wrote:
"with this script in the newest version of tok it takes 15 minutes for the full sample"
what is your pc CPU?
i have athlon xp 2000+ and takes only
7 minutes max to full sample with this script!

and:
"it takes 2 hours to do just the full sample in tok"
"Does this seem right "
8O
somethimg is wrong!
but the newest script is right.

:?

No, no way 8O, unless you have a Pentium or Celeron ~600Mhz or so.
Is that what you have :?:

-kwag

Hi,it is a celeron 900 but it is the same movie with the 2 different scripts so even if my computer is slow it did the first script in 15 minutes i just figured the 2 hours was a little too much.

Does the new script do a lot better job

Another thing i put the script into tmpge without doing any prediction and put the cq at 70 and it said about 13 hours which is what it usually takes for me to do a kvcd movie

Thanx

Just noticed that Kwag updated the latest script to include a little comment on Anime/cartoon sources. What do you guys think about Divx/Xvid sources? Is the optical script good enough, or should one add SpaceDust or some other filters? Thanks guys!

It does a great job on Divx sources, J-wo. I only use STMedian if the divx is a little noisy...

The Optimal Script For Anime/Cartoon material says:

Same as DVD clean material, but shorten the GOP to 12 frames (PAL) or 15 (NTSC)


Does this mean I would use 1-5825-2-1-15 instead of 1-5823-2-1-24 ?

and use 15 in file size prediction?

Thanks,

Andy

Exactly :wink:

-kwag

Has anyone been looking at reducing the macroblock artefacts
visible in divx and xvid sources? I see them in most divx & xvids
to some extent, mostly on areas of flat colour like walls. They're
very distracting when watching KVCD-from-divx on the TV!

I've had a little success by adding Blockbuster noise to the optimal
scripts, then increasing luma bluring and TemporalSmoother radius.
It doesnt seem to affect the quality of other parts of the picture much
but improves the look of the blocky artefacts.

Just wondering if anyone else has been poking around in this
area with any success...

Hi girv,

It's virtually impossible to remove visible macroblocks in a KVCD encode form DivX, if they are present in the DivX. Actually almost any DivX->KVCD will never be as good as a higher level conversion from DVD->KVCD.

-kwag

Thanks kwag. I know its gonna be hard to remove the macroblocks
as that would be equivalent to restoring the image data that xvid or
divx has quantised out (more or less). What I'm after is some way to
make them at least look better; my approach has been to try to
blend them away in areas of flat colour as that is where I notice them
the most, but I was asking if anyone had any other bright ideas to
improve the look :)

/girv

When possible, convert SVCD->KVCD rather than DivX/XViD->KVCD. The quality will be outstanding (assuming the svcd was done well and at a decent bitrate). I recently converted a SVCD cartoon using the modified GOP of 15 as Kwag suggested, and the result was the best cartoon conversion I've been able to do to date.

When stuck with Divx, your best best is to simply use the latest optimal script and bear with it ;) Most 2-CD Divx movies convert quite well to single CD KVCDs. The 1-CD Divx films will be quite a bit blockier.

Then I would use BlockBuster(noise method) with a high value, and disguise the macroblocks with uniform noise. This way it will look more natural. Try that :!:

-kwag

I know I always ask questions like this but... Do you have any hints on what settings I should use with BlockBuster noise in Moviestacker? And also where should I put it in my script in relation to the other filters? Thanks!

Use the default position where MovieStacker puts the filter. For "camouflaging" the blocks with noise :lol:, try this:
BlockBuster(method="noise", detail_min=1, detail_max=100, variance=0.8, seed=0 )
And increase with the "variance" value to obtain more "noise" on the picture.

-kwag