Avisynth: Update of Current Filter Combinations...
 

Hi all!

I'm just posting, to see what filter combinations you guys use at the moment! just post the script, and tell us what you think about it!

MergeChroma(blur(1.58))
MergeLuma(blur(0.05))
LegalClip()
GripCrop(480, 576)
GripSize()
Faerydust()
SmoothHiQ(5,25,35,200,10)
GripBorders()
DctFilter(1,1,1,1,1,.5,.5,0)
Letterbox(0,0,16,16)
LegalClip()

is what I use most of the time, works great for most films. If I want a sharper image I'll use SpaceDust() instead of FaeryDust, and Blockbuster(method="noise" variance=.6 seed=1) after SmoothHiQ if I'm getting too many DCT blocks.


EDiT: Source DivX/XviD AVI or DVD, although I normally won't need Blockbuster noise with DVDs.

oh, and remember..

Source from: LD; DVB; DVD; Capture AVI (DV vs. huffy vs. mjpeg vs. PIC) and VHS..

So we know why you used those filters.. he he..

-vhelp

I'm using avisynth 2.5 now. With both scripts i can get 2 films to dvd

For DVD Rips (i only go to DVD, not CD):

mpeg2source("test.d2v")
ac3Source("test AC3 T01 2_0ch 192Kbps DELAY 0ms.ac3")
#sampler(60) # - Use for any file prediction

AutoCrop(mode=0,wMultOf=16,hMultOf=16,leftAdd=16,r ightAdd=16,topAdd=16,bottomAdd=16)
Convolution3d (preset="movieHQ") #Or LQ if need more compression
DctFilter(1,1,1,1,1,.5,.5,0)
Limiter() #Limit color threshold to TV colourspace
ConvertToRGB24()

And For TV Caps:

AviSource("capture (2003-03-22 at 17-35-53).avi")
#FieldDeinterlace() #Use FieldDeinterlace if the source is interlaced
ConvertToYv12()
#sampler(60) #- Use for any file prediction
AutoCrop(mode=0,wMultOf=16,hMultOf=16,leftAdd=16,r ightAdd=16,topAdd=0,bottomAdd=0)
Convolution3d (preset="animeBQ")
TemporalCleaner()
UnDot()
DctFilter(1,1,1,1,1,0.5,0.5,0)
Limiter()
ConvertToRGB24()

I have a script like this for TV captures:

Fast and very good quality 8)

What template are you using Pyro??

I use this combination as others do and find it gives great quality and compression together. When using AviSynth 2.51 these are very fast filters.
Convolution3dyv12 (preset="movieLQ")
DctFilter(1,1,1,1,1,.5,.5,0)

KVCDx3 with the resolution dropped to 480x480/576, as my player does not support 528 or 544 resolutions. Actually I think that mergeluma should be around .4 instead of 8...I'll go edit my post.

I'm currently using:

On all my scripts :wink:

-kwag

do i need a special plugin to use mergechroma and mergeluma?

Nope. :)


If that's what you think is best, then I'll use it too! :D
...off I go to edit my post again...

I like 0.05 on luma, because even at ~0.1, it starts to blurr the image too much.

-kwag

Oh wait, 0.05...seems I misread it :?, I thought it was 0.5...maybe I'll stick to 0.4 as that looks fine to me...

well, if i add those two lines in my script, i get a message:

script error: there is no function MergeChroma


what must i add?

this is my script

Hmm, now that I've compared it a bit I see theres quite a difference...I suppose I'll go for 0.05 as well. So off I go to edit my post again. :lol:


Edit: Kane, try:

hmmm, nothing changed, still the same error

@ Kane,

May just be a Cut/Paste thing but...
It should read:
MergeChroma(blur(1.58))
MergeLuma(blur(0.05))

Your script reads:
MergeChroma(blur(1.58 ))
MergeLuma(blur(0.05 ))

Tighten your script...should work.

i got the mistake, i still had avisynth 2.07 installed, i just upgraded to 2.08 and it works :angel:

@ Kane,
Good to hear you got it.

I`m doing "Band of Brothers" Disc 1 as we speak..almost done.

Using:
Movie is 2hrs5min.
A mixed action movie, with muted colors...can`t wait to see how it turns out!

wow, with mergechroma and mergeluma one step of prediction lasts more than 30minutes 8O 8O 8O
but the video seems to be sharper with those filters.

Drop nomosmooth and fluxsmooth. See if it sky rockets in speed :wink:

-kwag

i just did , and the rocket is on it´s way :D

Ok, so you change the resolution in TMPGEnc to 480x480 but in your script you use GripCrop(480, 576) am I correct?

What does this mean "480x480/576" (the /576 sorry, I'm still a noob :lol: )

Aslo, should the resolution in your script match the resolution of the template?

if you live in a PAL region you will use 480x576, if you live in NTSC land , you use 480x480

it´s just the difference between the two standards

should i just leave nomosmooth away when doing file prediction, or can i completely drop it, if i use mergechroma and mergeluma

I'm not using nomo or flux anymore. I just use mergechroma+mergeluma+spacedust+C3D.
If it's a noisy source, then I use faerydust instead of spacedust.
The difference in file size, is not that much smaller by adding flux, nomo, or other filters. At least, that's what I've been able to experience.
Even SansGrip agreed once, that space + C3D was plenty. With the added bonus of merging chroma and luma, it's good enough :D

-kwag

I just use mergechroma+mergeluma+spacedust+C3D.


that´s exactly my script http://www.digitalfaq.com/archives/error.gif

the sample looks great. now i will let ToK do the work and see what comes out

If the sample looks great, then the full encode will also look great :D

-kwag

What about TemporalCleaner, should it be dropped as well? (just like nomo and flux) if MergeChrom & MergeLuma & Spacedust were used

You might want to keep TemporalCleaner. It does reduce about ~80-100KB per minute, depending on the source. DVD sources are pretty clean, so it doesn't make that much difference. Leave it in there, as it's a very fast filter, and it won't slow down your encode :wink:

-kwag

:?: :?:

Yes. Otherwise you will be resizing with TMPGEnc.

Ok, let me see if I've got this correct. If I have a regular DVD movie rip, which I believe has a resolution of 720x480 NTSC, and use the KVCDx3 template which has a resolution of 528x480, I should put in my script GripCrop(width=528, height=480) right?

Now if I want to resize the movie, I would do so in TMPEGEnc but leave my script as is or change it according to what I resized the movie to in TMPGEnc?

The final resolution in TMPGEnc is what determines if a movie will play in a player, not the resolution in the script right? Or is it both (resolution in TMPGEnc and script)?

Yes.

You should use the same size in your GripCrop line as you use in TMPG. Otherwise you are doing unnecessary resizing with TMPG (TMPG is the final step in the video-creation process, so if you were to use GripCrop(width=528, height=480) in your avisynth script, but set the sizing in TMPG to 352x480 for example, you'd end up with an m1v file which is 352x480). counter-productive.

True, but you should use the same sizing in both for best results.

Regards,
-d&c

this combination is great!!!!!
i got the best results since i work with kvcd templates

Encoding too long,which filter really useful?
 

Hi I'm a newbie, so please excuse my question.
First I want to praise you all for this great work.
I want to encode a divx-avi to mpeg1 528*576 kvcd 102min on 1cd, bitrate 100-2500.
This is my script with recommendations from kwag (mergechroma and luma):

AviSource("V:\Monty_Python_Der_Sinn_des_Lebens.avi ")
ConvertToYUY2()
LegalClip()
LanczosResize(496, 384, 2, 7, 636, 346)
mergechroma(blur(1.58))
mergeluma(blur(0.05))
SpaceDust()
TemporalCleaner()
Convolution3D(preset="movieHQ")
AddBorders(16, 96, 16, 96)
LegalClip()

But encoding with tok and tmpenc with cq63 after prediction
takes too long: 25h.
So which filter should I disable first to decrease time? mergechroma,mergeluma,SpaceDust,TemporalCleaner or Convolution3D?
Or bilinearresize/bicubic instead of LanczosResize?
I don't need highest quality as possible, but I love sharp pictures.
Could anybody help me please.

Second question: where can I find avs script templates for different sources (vhs, dvd...). I have been serching for the last 3hours.
Thank You!

Hi feldi,

First of all, what processor/speed do you have :?:
For different filters, you can download MovieStacker or SwiftAVS, and there you can select different filters that you want. For example, different presets for Convolution3D, etc :)

-kwag

good speed quality compromise?
 

Only 700mhz AMD Duron - I know it's not the best.
Now I don't use C3D and it's much faster: about 15h.
What do you think, is this the best compromise?
Should I always use mergechroma and blur (with the settings in my script).
Is my script ok (made with movie stacker)?
Is lanzsos-resize the best choice?

Too many questions? -please forgive me.

Thanks, feldi

Hi feldi,

Your processor speed is not the best option for encoding :roll:
As you did, removing the slowest filters will increase speed, at the cost of quality :!:
Try Bicubic resize instead of Lanczos. It's not as sharp as lanczos, but it is sharper that bilinear.
Your script is good :wink:

-kwag

[quote="kwag"]I'm not using nomo or flux anymore. I just use mergechroma+mergeluma+spacedust+C3D.
If it's a noisy source, then I use faerydust instead of spacedust.
My source is kind of "noisy" because it's not DVD, it's SVCD. On top of that, my SVCD source is split into 2 cds and I want to squeeze them into one with KVCD settings. I do notice some blockiness, visible in the out of focus backgrounds and in light gradients (like spotlights or a sunbeam).

So what filters would you add on top of that? Would you go up to pixiedust? And I know you've said many times that Blockbuster isn't needed at KVCDx3 resolutions, but because my source is SVCD I'm encoding at 480x480. So I'd love to hear what filters/settings I should try. Thanks!