What are all these avisynth plugins for?

[rkorkie]

I have noticed over the past while that people are using all sorts of plugins and filters to create their DVD content. I, for one, simply follow the robshot method with CCE 2.50 and leave it that. However, I noticed that when I compare frame-by-frame from the original DVD-9 to the CCE output file, the quality isn't quite the same. Maybe a bit more blurry, not quite as sharp. I noticed this on the Matrix last night.

What do all the plugins and filters accomplish, and are they really more overhead than they are useful? I notice a lot of talk about the Adobe Premiere plug-in with avisynth, but is that just for video captures using the premiere software?

I am looking just to transcode DVD-9's to DVD-R, and I have noticed that a lot of people's .avs scripts contain more than just the basic:
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
mpeg2source("D:\rmd\rmd.d2v")
ResampleAudio(44100)

Any input into what all these plugins and filters do would help a lot. Thanx.

[Asmodian]

Umm, read through the threads in this forum. I find the filters are useful for making the video look better using all sorts of techniques. (smoothing, sharpening, resizing, improving compressibility, deinterlacing/inversetelecine, many more)
I do go to xvid rather then mpeg2 but I believe most of the gains for encodes destined for xvid also help when going to mpeg2. My encodes do take ~3-4 times longer then they would with out the filters :)

I usually use

Telecide - reconstructs the progressive video stream (Decomb)
Decimate - Decimates ;) (also from Decomb)
Convolution3d - Improves compressibility or smoothes depending on settings
Resize - you might not want this so as to keep DVD specs

[rkorkie]

I have read through the threads in the forums, and it's clear to me that people do indeed use filters, but it's not clear to me as to why or how. It seems that everyone is encoding video for a slightly different purpose, but mine is just the basic dvd transcoding to fit on to a dvd-r.

How come doom's guides don't mention the usage of any filters? I'm still stuck...

[Asmodian]

I am confused, what do you want to know exactly? All the plugins say what they do and how to use them in their documentation. I would recommend using convolution3d which smoothes without loosing very much detail to improve compressibility.

I usually use the line convolution3D(0,6,8,6,8,2.8,0) in the avs before a resize (you would do it last because you don't resize). This might over smooth for you so you could try 0,4,6,4,6,2.8,0 there are lots of options for you to choose, what do you want to do and what types of DVD's are you transcoding (NTSC PAL)? There is also a version of mpeg2dec (mpeg2dec3 from MarcFD avalible here http://ziquash.chez.tiscali.fr/ ) that has post(pre?)-processing to remove blocks and ringing from the source. Most filters are fixing some problem in the source (telecined, noisy, levels problems, etc) if your source is of great quality they you might not need many/any filters.

Is this the kind of help you wanted?

ps. inverse telecine is the done with the combo of telecide and decimate (there are other methods) and reduces 30fps video to 24fps by removing the frames the studio added when going from 24(film) to 30(NTSC DVD) - this improves quality by reducing the total number of frames to save by 20%.

Edit: Doom9's guides do not mention filters because they update very very often and are hard to use (most don't have settings that just work for everything and when used wronge many cause more harm then good) convolution3d may be an exception but it is very new, maybe he will include it at some point in the future :D

[Acaila]

I believe most people posting about filter usage in this section transcode from DVD to a format with constricted filesize. People encoding to DivX/XviD/VCD/SVCD/XVCD etc. all have much less available space compared to the original DVD. So most people need filters to smooth/resize or otherwise change the original in order to still retain acceptable quality in the limited amount of space they have.
If you're just copying the original DVD to another DVD-R you usually don't have a filesize limitation and therefor don't need any filters to "improve" the picture.

[Asmodian]

I just thought that when going from a DVD-9 to a DVD-R one transcodes to drop the bitrate (so it will fit on the single layer disc) so maybe a light convolution3d or post-processing (with mpeg2dec3) would help. The post-processing also might help with not reencodeing the compression artifacts from the original DVD (I usually can notice artifacts on DVD's). I Don't really know though, having never done a DVD-9 to DVD-R transcode :)

[rkorkie]

Thanx for the replies. Yes, I'm not really looking to improve compressibility that much, given that the longest movie I'll ever encode is about 3 hours, and when using CCE's robshot, you don't have to sacrifice much to get it onto a DVD-R. However, with the compression that is occurring, I was looking for whatever filters or plug-ins were the best bang for the buck (ie. best improvement vs. time overhead). Does that make my intentions more clear and specific?

I did do some reading last night, and people are doing 4 or 5 VBR passes with CCE, whereas I was only doing one. Could that perhaps be the fix, or would one of the many filters do a better job in less time?

Thanx again.

[Asmodian]

I would say you should go for the multiple passes over filters, most of the filters are for fixing problems with the source, improving compressibility, or sharpening/smoothing - there really isn't a fast 'make the video look better' filter. Having said that there are a few fast (relative) sharpening filters you might want to try (you said your encodes looked a little blurrier then you would have liked?). I should mention that I have never used CCE so should really leave this question for those more experienced with it.

[theReal]

The only filter I'd maybe use for a high quality DVD to DVD conversion is convolution3d. If the source DVD isn't perfect (which it isn't quite often) then C3D using low settings can really make it look better. However, it isn't really fast...
All other smoothing filters would be overkill for your purposes and make the encoding look worse than without, I think.

[rkorkie]

Thanx for the advice guys. I downloaded convolution3d, put it in my avs file and then into CCE 2.50, and it crashed CCE immediately. This is my file:

LoadPlugin("C:\DVD\MPEG2DEC.dll")
mpeg2source("C:\DVD\matrix.d2v")
Convolution3D (0,3,4,3,4,2.8,0)
ResampleAudio(44100)

I put convolution3d.dll in the c:\dvd directory. Am I missing something here?

[rkorkie]

ok, scrap that last post, I got it figured out. I'm gonna test convolution3d with these settings:

LoadPlugin("C:\DVD\MPEG2DEC.dll")
LoadPlugin("C:\DVD\convolution3d.dll")
mpeg2source("C:\DVD\matrix.d2v")
Convolution3d (preset="movieHQ")
ResampleAudio(44100)

If I find it's better than without a filter, I'll post that too...of course, this filter doubles the encoding time approximately.