Best AVS Script

[Sigmatador]

for a dvdrip my avs script are very simple (crop resize and sometimes deinterlace ivtc).
but some test with scripts founded on this board show me that the visual quality and the compressibility can be improved.

so can you tell me some good scripts for MovieDVDRip and AnimeDVDRip ?

[WarpEnterprises]

It's very important to know:
- which resolution (e.g. VCD 352x288)
- which quality/bitrate (e.g. 50min or 100min on 1 CD)
- which compression (DIVX, VCD, SVCD, DVD)

Tell us!

[Sigmatador]

oops sorry ^^
to compress with XviD for a 1cd rip
and for any resolution (choose one for an example, it's easy to change ^^) but 576*xxx or 640*xxx for 90 to 120 min (for movies) and sometimes more for anime (150min)

[Marc FD]

....... good luck .........

[Sigmatador]

:confused:
maybe i have to use copysame for anime

[manono]

Hi-

but 576*xxx or 640*xxx for 90 to 120 min (for movies) and sometimes more for anime (150min).

If I understand you correctly, you're implying that anime is more compressible than live action movies, but it's the opposite. I'm with Marc FD-Good Luck.

[Sigmatador]

arf, yes i was thinking that anime was more compresible than movies.
so forget the more time for anime.
it's possible to code a 90-120min film with xvid for 1cd no ? (i think i don't need luck for that ;) )
i read here that a good avscript (using Copysame, convolution3d, cnr2, temporalsoften, temporalsmoother, MAM, SmoothHiQ, MergeChroma) can improve quality and compressibility (for xvid 1cd rip) and i want to know some good scripts.
can somebody help me ?
i said "for movies and animes" because i thinked to having a good improvement, the avs script should be different for anime and movie.

[theReal]

Just try convolution3d with the default settings (the ones recommended by the author in the included readme). This filter improves compressibility really much without giving the video a "filtered" look.
I just used it for a tv-capture and I like it better than anything else I've tried before. I don't know what it looks like for DVD sources, but I can imagine it's quite to improve compressibility.

I usually encode DVD's without filtering, I rather use one more CD - the results are better, usually. However, if I have to use a filter on bad DVD (hard to compress), I'd try convolution3d, or temporalsmoother(2,1) (up to (4,1)).

I don't encode anime, but usually you can use stronger spatial filters on drawings (or other filters, like warpsharp, that are not good for normal movies)

[Didée]

theReal,

one comment, regarding WarpSharp:

I use it (with lower amount) for most of my TV captures. My captures are of good quality, but a little to "soft" for my taste.
To compensate for this, IMHO a little(!) WarpSharping is ideal - every other sharping method increases artefacts.
As a side-effect, throwing in some WarpSharp also denoises and increases compressability - both not by too much, but mentionable.

If only someone ported WarpSharp to process YUY2 ...

[Koepi]

I'm trying the matrix at 720x304 with convolution3d in some decent settings for "SCD"

Convolution3d (0, 8, 8, 8, 8, 0)

The pictures i watched in preview were really nice, I didn't realize loss in detail, and the noise is gone.

Compressability increased from 1.7GB first pass to 1.2GB (still ~5 minutes to go but I don't expect something bad to happen anymore ;) )

I use MVH, so the 2nd pass should be somewhat faster than 10 hours.

I'll report back then.

Best regards,
Koepi

[iago]

@Koepi

So nice to hear that :). I'm looking forward to see your results with convolution3d with these settings ;).

regards,
iago

[Defiler]

There's a readme file for Convolution3D?? Where?

[Sigmatador]

ok i'm going to test convolution3d with this setting to (on matrix ^^)

[vlad59]

Originally posted by Koepi
I'm trying the matrix at 720x304 with convolution3d in some decent settings for "SCD"

Convolution3d (0, 8, 8, 8, 8, 0)

The pictures i watched in preview were really nice, I didn't realize loss in detail, and the noise is gone.

Compressability increased from 1.7GB first pass to 1.2GB (still ~5 minutes to go but I don't expect something bad to happen anymore ;) )

I use MVH, so the 2nd pass should be somewhat faster than 10 hours.

I'll report back then.

Best regards,
Koepi

I'm glad you like it.
I'm impressed by the compressibility increase with the _soft_ settings you used (1.7Gb to 1.2Gb is huge difference). I think you won't see any detail loss (Conovolution3D (0, 8, 8, 8, 8, 0) should keep a lot of details) except maybe in fade in or fade out scene : this was the only kind of scene where Convolution3D may sometimes create a too sharp frame (harder to compress). I don't remember if there is such a scene in Matrix ???

I'm waiting for your report.

@all

My filter is only available in the avisynth thread : Smooth in time and space .
Sorry but I don't have a website and I don't like HTML :devil: at all.

[vlad59]

Sorry to pollute this thread with the following question ;)

@Koepi

Iago in an other thread seem to have avisynth crash when using Convolution3D in second pass.
Are you having the same problem ?
What version of avisynth are you using ?

Thanks in advance.

[theReal]

There's a readme file for Convolution3D?? Where?

Defiler, in the latest beta (08/17/02) there's a readme file. OK, it's not called "readme", it's called "convolution3d.txt" :)

vlad59, good work!! The compressibility increase with your filter is indeed better than with any other smoothing filter I have tried so far (even combinations of chroma noise reduction and temporalsmoother don't achieve that level - at least I can say so for tv captures)

[Koepi]

Vlad,

i don't experience crashes or segfaults with convolution3d, but I use avisynth 2.0.2...

The only artefact I've seen so long is some ghosting on a scene change - from a bright scene/picture to a dark scene/picture, there were clear artefacts visible.

For tv captures convolution3d is the best yet :)
Well, I'm just testing on DVD, but the encoding still needs 6 hours...

Best regards,
Koepi

[Sigmatador]

which scene have the ghost effect ?

[vlad59]

Originally posted by Koepi
Vlad,

i don't experience crashes or segfaults with convolution3d, but I use avisynth 2.0.2...

The only artefact I've seen so long is some ghosting on a scene change - from a bright scene/picture to a dark scene/picture, there were clear artefacts visible.



Ghosting ! With Convolution3d (x, x, x, 8, 8, x), that must a very special case.
Is this a Matrix scene ?? I own this DVD (R2) so I can make a test if you say me how to find this scene.

@TheReal and others
Thanks for giving a chance to this filter even if it's so slow. ;) ;)

[Sigmatador]

Thanks for giving a chance to this filter even if it's so slow

i don't care if it's slow ^^. my priority: quality/compressibility :cool:

my tests aren't finished, but your filter may be good.
but... i have to check there is no ghost effect ;)

[kyousuke]

to every body, do you use the vdub version or a dll compiled for avs ?

i don't find any dll ..... :/

thanks

[theReal]

Thanks for giving a chance to this filter even if it's so slow.
The speed is not so bad, a combination of Spatialsoften and Temporalsmoother is slower than Convolution3d. Even Spatialsoften alone (the newest version from avisynth 2.03) is about the same. But the effect of convolution3d is better and it's good for very subtle filtering as well as very strong filtering.

[iago]

Hello everybody,

Since Koepi is also doing a test encode with the same filter settings, I find it suitable to discuss my test results with convolution3d(0,8,8,8,8,0) in this thread too.

The movie is the same again: The Blues Brothers (1980), the test encode results of which with different filter settings were discussed in http://forum.doom9.org/showthread.php?s=&threadid=31888.

Since vlad59 also remarks that he hasn't tested the filter with normal films a lot, I hope such tests will help finding some suitable convolution3d settings for "normal" movies too.

To repeat:

TEST SETUP

Movie: The Blues Brothers (1980) - NTSC 1
2hr-27min/512*272/SimpleResize
Aimed Video Size: 650000kb (635mb)

SmartRipper 2.40
DVD2AVI 1.76 (ForcedFILM 29.97 -> 23.976)
Gordian Knot 0.26 beta (crop/resize)
VirtualDub 1.4.10 (fast recompress)

XviD 2pass
Koepi's 18082002-1 build
1stPass: 6.UltraHigh/H.263/Lumi enabled
2ndPass: 6.UltraHigh/MPEG/Lumi enabled/Min-Max I-frame:2-6/Min-Max P-frame:2-16/default Alt.CC/all the rest default
Maximum-Minimum I-frame interval: 250-1 (both passes)
Hinted ME

Script (Avisynth 2.04+/15082002):
-------------------------------------------------
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
LoadPlugin("C:\PROGRA~1\GORDIA~1\SimpleResize.dll")
LoadPlugin("C:\PROGRA~1\GORDIA~1\convolution3d.dll")
MPEG2source("D:\BLUSBRO\BLUSBRO.d2v").crop(3,7,714,460).SimpleResize(512,272)
Convolution3d(0,8,8,8,8,0)
Trim(0,205210)+Trim(205211,0).TemporalSmoother(10)
--------------------------------------------------

I'll post the results when first and second passes are finished.

best wishes,
iago

[Sigmatador]

oki i'm going to help testing.
find a best convolution3d for most movies is a very good idea
(coding austinpower2 ^^)

[iago]

Hello everybody,

Here are the 1stpass and 2ndpass results:

Speed: ~ -25% slower than unfiltered encoding / quite acceptable imho
Compressibility: 1stpass/2ndpass ratio: 2.42 (which means great increase in compressibility even with these non-aggressive settings / it was 3.93 unfiltered)
PS: no crash with avisynth 2.04+ (15082002)

2ndpass results:
-----------------------------------------------------
DebugView analyzer for XviD codec v0.11 by MoonWalker
e-mail : s_ilias@gmx.net

Quantizers Analisis
---------------------

Quantizers Used For Movie :
------------------------------
Quant 2 Used : 192 Times, Percentage Used : 0.09%
Quant 3 Used : 9663 Times, Percentage Used : 4.71%
Quant 4 Used : 78705 Times, Percentage Used : 38.35%
Quant 5 Used : 82000 Times, Percentage Used : 39.96%
Quant 6 Used : 30676 Times, Percentage Used : 14.95%
Quant 7 Used : 3720 Times, Percentage Used : 1.81%
Quant 8 Used : 235 Times, Percentage Used : 0.11%
Quant 9 Used : 11 Times, Percentage Used : 0.01%
Quant 10 Used : 1 Times, Percentage Used : 0.00%

Average Quantizer Used for Movie : 4.709

Quantizers Used For Credits :
--------------------------------
Quant 31 Used : 7333 Times.

MPEG Quantization Type Used 17188 timed, Percentage Used : 8.09%

H.263 Quantization Type Used 195348 timed, Percentage Used : 91.91%

Quantizers prevented from rising too steeply 123 times


Intra-Frame (Key-Frame) Quantizers
------------------------------------

Movie
-------
Quant 4 Used : 2135 Times, Percentage Used : 73.88%
Quant 5 Used : 19 Times, Percentage Used : 0.66%
Quant 6 Used : 71 Times, Percentage Used : 2.46%

Credits
---------
Quant 31 Used : 665 Times, Percentage Used : 23.01%

Number Of Consecutive I-Frames : 621

Inter-Frame (P-Frame) Quantizers
------------------------------------

Movie
-------
Quant 2 Used : 192 Times, Percentage Used : 0.09%
Quant 3 Used : 9663 Times, Percentage Used : 4.61%
Quant 4 Used : 76570 Times, Percentage Used : 36.52%
Quant 5 Used : 81981 Times, Percentage Used : 39.10%
Quant 6 Used : 30605 Times, Percentage Used : 14.60%
Quant 7 Used : 3720 Times, Percentage Used : 1.77%
Quant 8 Used : 235 Times, Percentage Used : 0.11%
Quant 9 Used : 11 Times, Percentage Used : 0.01%
Quant 10 Used : 1 Times, Percentage Used : 0.00%

Credits
---------
Quant 31 Used : 6668 Times, Percentage Used : 3.18%


Frame Analisis
----------------

Number Of Intra-Frames (Key-Frames) : 2890
Number Of Inter-Frames (P-Frames) : 209646
Total Number Of Frames : 212535

1.36% of the Movie is Intra-Frames (Key-Frames)
98.64% of the Movie is Inter-Frames (P-Frames)


Size Analysis
----------------

1-Pass Size : 1618023266 Bytes or 1580100 KBytes
Scaled Size : 660989388 Bytes or 645497 KBytes
Actual Size : 660832694 Bytes or 645344 KBytes

Usefull Statistics
------------------
Compressibility : 40.84%
Relative Quality of XviD avi : 42.47%
Absolute Quality of XviD avi : 91.87%
------------------------------------------------------

Finally a few comments: compared with the previously tested filters/filter chains with the same source (see http://forum.doom9.org/showthread.php?s=&threadid=31888), convolution3d(0,8,8,8,8,0) gave the best result so far, though still being far from a decent final encode. But here I want to emphasize one more time that this was a real hard-to-compress case and the average quantizer 4.709 is too high a value to reach decent quality imho. Therefore it would be much more accurate to test this convolution3d setting with better compressible movies before making a final comment.

best regards,
iago

[Sigmatador]

good job ^^
can you make screenshot to compare with your other test (with dbr2 an MAM) plz?

[iago]

@Sigmatador

Sorry but I didn't take screenshots this time and didn't keep the encoded file (which as I said was still far from decent quality) after I wrote my comments on it.

iago