Size totally off!!

[AcidJunkie]

yesterday i ripped my first movie with XviD (latest koepi). the movie was "Jason goes to Hell" (PAL, 4:3, 86:40 min, little noise).
it was only 291 MB big ;(
what have i done (totally) wrong?! Plz help.

avs:
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
LoadPlugin("C:\PROGRA~1\GORDIA~1\Convolution3D.dll")
mpeg2source("C:\PAYBACK\jason9.d2v")
crop(3,0,709,573)
LanczosResize(528,400)
Convolution3D(0,4,4,4,4,2.8,0)

here my settings: (audio ogg (=> final file size 52 MB)
1st pass load default setting
6 - ultra high
h.263
xvid
300 max-i intervall
1 min-i intervall
discard firstpass
hinted me
(credits seperately)

2nd pass
6 - ultra high
new modulated hq
xvid
300 max-iintervall
1 min-intervall

Min I 2
Max I 4
Min P 5
Max P 15
high bit scenes: 0%
low bit scenes: 10%
payback delay: 300
low curve aggression
150 High D
75 Low D
50 % Strength
2500 Max Bitrate

Desired size (Kbytes): 660479

want to do another 2nd pass only if i have done nothing wrong up to 2nd pass tonight
ACID

[AndyP]

I may be barking up the wrong tree but min p should not be 5 it should be 2 and why have you changed max bitrate??

Andy

[AcidJunkie]

i thought max bitrate should be the max bitrate allowed to use for a frame. should i leave it at 10000?
min P read here http://forum.doom9.org/showthread.php?s=&threadid=28947&highlight=515
(koepies reply from 11th july 2002 16:47)

ACID

[Lefungus]

->new modulated hq
->Min I 2
->Max I 4
->Min P 5
->2500 Max Bitrate
->Convolution3D(0,4,4,4,4,2.8,0)

As this is your first movie ripped with xvid, you should consider using the button "load defaults" in xvid configuration tab. Get used to "standard" settings first, then try custom settings

Idem for avisynth plugins like convolution3d. Even if it's good, try without first

[AcidJunkie]

i´ve read a lot yesterday before testing these settings...
but maybe you're right. i leave the convolution3d filter and do another first pass with standard settings right now.

ACID

[AndyP]

Personally I use I min-max as 2-4 and P min-max as 2-8 and pick a resolution that ends up with average quant of 3ish (corresponds roughly to a DivX 5 comp. test in GKnot of 80%). I am almost certain that with min P as 5 you will get a bad encode.

I have never played with max bitrate as I really don't understand it.

I might be wrong but your first pass is probably OK.

Kind Regards,
Andy

PS did you use koepi's stable or dev build??

[AcidJunkie]

maybe 1st pass is okay, but i make another one without convolution3d filter. so i can do 2 2nd passes, one with & one without convolution3d in order to see the differences. but i'm quite sure that convolution3d is good, esdpecially for this movie (litlle noise, 4:3).
i`m new to convolution3d... i used temporal smoother several times in noisy movies.

ACID

PS i used stable build (04102002)

[manono]

Hi-

AndyP is right, AcidJunkie. Koepi was referring to something completely different with his 5-15 remark. Too much of the Acid was clouding your brain, I think.:) Using minimum P-Frame of 5 also explains your undersized movie, and it probably didn't look so good, either. AndyP's quant recommendations are good. Good Luck on the next try.

[AcidJunkie]

thx for the fast help guys... will try it on my laptop, too and keep you informed with the results...

ACID

[AcidJunkie]

yep. AndyP is right - i mixed some things up (maybe it was too late). setting min/maxP to 5/15 is total nonsense. i read something more deeply through the forum yesterday. so i left the standard settings (load default).
i did the 1pass & 2pass without any filter used and lanczos resize.
i changed nothing besides 6-ultra high, hinted ME, and credits settings...
the result seemed to be okay for me (since it is a noisy source) and as i didn't change anything. besides i wanted to figure out changes and its effects in comparison to standard settings.
the other 2nd pass i choosed same settings besides MaxI-Intervall: 250, MinI-Intervall:5, Payback Delay 300 (next time i will set it even higher). HighD 200, LowD 90 (next time 80) (resulting from PerfectXviD), Strenght 50% (seems to work well)...
And I'm REALLY happy with the results!!
Now I'm going to do the second procedure again this time using Convolution3D with settings as in my above post. (it takes double encoding time-is that normal?).
I'm letting you know the results later...

thx
ACID

[AndyP]

Yeah, Convolution slows things down for me as well. But it is well worth it. I use the same settings as you said at the start (0,4,4,4,4,2.8,0) but if you have a noisy source then (0,8,8,8,8,3,0) might be better to start. Look in the avisynth forum and there are lots of different settings suggested.

As far as the AltCC settings go, I don't bother to change from default anymore as I could never really improve (or even see a difference). Sometimes using Koepi's statsreader and using external CC I can get a better quant distribtuion (but not always). The only settings I change are to use MPEG for both passes (I prefer it to H263) and capping quants as described. I don't use lumi or MVH.

Some are having successes with pre-processing with MarcFD's modified mpeg2dec dlls. I haven't tried them yet.

I have been using xvid alot recently and am now convinced that it almost always beats divx5. The thing that convinced me was in Independence day (the scenes in area 51 when they are dissecting the alien and it decides to wake up) there are lots of flashing lights and divx5 just dissolves into a block-fest while xvid handles it perfectly!!! :D :D

Kind Regards,
Andy

[AcidJunkie]

i'm going to test the (0,8,8,8,8,3,0) setting, too.
mpeg setting over h263 i mostly read not to use it for 1cd rips. and i prefere 1cd rips. but i will test it, too.

so far & thx
ACID

[AcidJunkie]

having done a compressability with divx5 in gknot (15%) i have 79.2% of 0.198 at 592*448 (0.157 Bitts/Pixel*Frame) - which is very high i think for a 1cd rip (4:3).

*EDIT*
ok. mixed something up with compressability check for divx5 with LanczosResize and Convolution3D used. so i choose again 528*400.

nevertheless i will test it now with convolution3d (0,8,8,8,8,3,0) as AndyP prposed since it is a bit noisy source. and my firts setting (0,4,4,4,4,2.8,0) is more a setting for normal or clean sources.

*EDIT*
i used (0,6,10,6,8,2.8,0)

the other settings (see above) in my 2nd pass seemed to work well for that movie. besides the highd and lowd settings (perfectdivx).
with that high resolution i don't have to care about using a sharpen filter afterwards (i mean if the convolution settings are too high).

*EDIT*
returned good results - i will leave it so
thx
ACID