Log in

View Full Version : Really BIG undersize using Koepi xvid 1.0 beta 3


Pages : [1] 2

split710it
7th January 2004, 09:35
First at all, I "searched" before I found many thread but not one as this and any about new koepi 1.0 beta 3

Koepi Xvid ver 1.0 beta 3

Hi

I encoded from MPEG SVCD a movie with original starting size 1.600.000 Kb (1.600 mega), 480x480, 29.97, NTSC.

My desired final file is: 544x288, IVTC to 23,976, cropped

two pass
B frames = 2
Quantiz. = MPEG

on second pass i setted final desired size for 1.249.000 KB (1.249 mega).

used gknot to prepare the avisynth AVS file and virtualdub (fast recompress) to encode.

Virtualdub 1.5.10 did not give me mistakes or something strange but at the end resulted ONLY 168.000 KB size for new AVI (with right new resolution and new framerate 23.976)

I can understund Undersize... but... there is a BIG UNDERSIZE in that case (160.000 instead 1.249.000) ??

I saw new specification of beta 3 (fixed problem with bitg files)

Re-encoded with usual Xvid koepi "XviD-Koepi24062003" size resulting is exactly 1.249.000

could some explane me bettere what is happened???

ciao

Koepi
7th January 2004, 09:57
Hit "load defaults" after installing a new xvid build.

Then do the encode again.

Ah wait. Did you read syskin's signature? - all programs that automatically configure xvid might not work with 1.0 tree (including gknot)

Next time do a little research if you're doing things properly please :)

Koepi

split710it
7th January 2004, 10:18
Thank's a lot for your answer...

You mean that problem is only concerning Gordian knot using???

all programs that automatically configure xvid might not work with 1.0 tree (including gknot)

Uhm... i used it only to compile avisynth file, (DVD2AVI, saved a D2V project, calculate size, resize, saved ONLY with IVTC). I just used only AVS + VD configuring manually xvid codec.

ciao

split710it
7th January 2004, 10:36
This is my AVS: (I Just only changed now manually movie title with XXXX in source for privacy, on the original there is the right one wrote)..

#
# Creato con Gordian Knot
#
# http://gknot.doom9.org
#
# PLUGINS
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec3.dll")
LoadPlugin("C:\PROGRA~1\GORDIA~1\undot.dll")
LoadPlugin("C:\PROGRA~1\GORDIA~1\decomb.dll")
#LoadPlugin("C:\PROGRA~1\GORDIA~1\KernelDeInt.dll")
#LoadPlugin("C:\PROGRA~1\GORDIA~1\dgbob.dll")
#LoadPlugin("C:\PROGRA~1\GORDIA~1\Convolution3d.dll")
#LoadPlugin("C:\PROGRA~1\GORDIA~1\FluxSmooth.dll")
#LoadPlugin("C:\PROGRA~1\GORDIA~1\TomsMoComp.dll")
#LoadPlugin("C:\PROGRA~1\GORDIA~1\VSFilter.dll")
#LoadPlugin("C:\PROGRA~1\GORDIA~1\SimpleResize.dll")
#
# SOURCE
mpeg2source("F:\a\XXXXXXX.d2v")
#
# TRIM
#trim(startframe,endframe)
#
# IVTC
Telecide(order=1,guide=1).Decimate()
# o usa
#IVTC(44,11,95)
#GreedyHMA(1,0,4,0,0,0,0,0)
#
# CROPPING
crop(0,84,480,308)
#
# DEINTERLACING (1)
#FieldDeinterlace()
#FieldDeinterlace(blend=false)
#TomsMoComp(1,5,1)
#
# DEINTERLACING (2)
#KernelDeInt(order=1,sharp=true)
# o forse
#DGBob(order=1,mode=0)
# DEINTERLACING (3) - richieste speciali
#GreedyHMA(1,0,0,0,0,0,0,0)
#Telecide()
#SeparateFields()
#
#
# SOTTOTITOLI
#VobSub("FileName")
#
# RESIZING
BicubicResize(544,288,0,0.5)
#
Undot()
#
# DENOISING: scegli una combinazione (o nessuna)
# 1) poco noise
#Temporalsoften(2,3,3,mode=2,scenechange=6)
#mergechroma(blur(1.3))
#FluxSmooth(5,7)
#
# 2) Medio noise
#Temporalsoften(3,5,5,mode=2,scenechange=10)
#Convolution3d("moviehq")
#FluxSmooth(7,7)
#
# 3) Tanto noise
#Temporalsoften(4,8,8,mode=2,scenechange=10)
#Convolution3d("movielq")
#FluxSmooth(10,15)
#
# BORDI
#AddBorders(left,top,right,bottom)
#
# CONTROLLO DI COMPRIMIBILITA'
# !!!!Dimens.Ritaglio ora dev'essere di 14 per poter essere usato con GKnot!
#SelectRangeEvery(280,14)
#
# FOOL CCEnc
#empty = BlankClip()
#AudioDub(last,empty)



I thougnt it is a normal AVS script.... where is the wrong part.. is it the D2V extension (tipical for DVD2AVI)

ciao

digitalman
7th January 2004, 13:41
I have the same issue using the past two builds of XVID with latest version of Avisynth. VirtualDub or VirtualDubMod is being used, not GordianKnot or any other automatic program out there. It seems to only occur when using Avisynth with DVD2AVI. This occurs on any type of movie, such as high action, low action, or dark scene movies.
When I use VirtualDubMod and load the MPEG directly without Avisynth, the undersize does not occur.
-I am using Donald Graft's version of DVD2AVI and MPEG2DEC3.DLL located at www.neuron2.net.
-I also used the same Avisynth setup with another codec I will not mention, and the undersize did not occur with that.

Mango Madness
7th January 2004, 16:30
sounds like you need to format your harddrive, too much crap from gknot left over to b0rk things up (just my opinion). i have no problems at all with my encoding process done manually.

duartix
7th January 2004, 17:00
Just checking:
What was your first pass size?
Are you sure you haven't maxed out the codec?
With your settings it's not that dificult to happen...

Maurizio
7th January 2004, 17:21
Dont' use DVD2AVI. it has some problems.

Use simply MPEG2DEC3. it reads directly the vobs and creates
automatically the d2v file.

note that now it's kept up to date by Nic.
Look at
http://nic.dnsalias.com/mpeg2dec3.html

Also, use AVS2AVI, it's simpler and quicker.
it's here :
http://daveel.leffe.dnsalias.com/
Regards
Maurizio

split710it
7th January 2004, 18:00
I think there is some confusion....

1 - is not possible to set size during first pass (in all xvid codec)
2 - I made already some Xvid in many ways
DVD to XviD, SVCD(MPG) to XviD, M2V(SVCD) to XviD, Divx to XviD

I always "ONLY" used Gordian Knot as a gui to write the script, I always encoded "MANUALLY" using VD or VDmod

DVD2AVI to prepare the D2V after that only Bitrate and Resize sections
at the and to save invcluding IVTC (when file is in 29.97) in AVS.

after that Gordian Know it doesn't exist anymore for me....

Now
This is the first time I have that problem, and with same AVS size will create WELL using versions of Xvid codec released before new 1.x

Tried with the same AVS too..

ciao from italian friend

mikeX
7th January 2004, 18:23
1 - is not possible to set size during first pass (in all xvid codec)
what duartix means is what was the size of the video produced by the 1st pass (unchecking the 'Discard first pass' box somewhere in the xvid 1 configuration window will do that)

an alternative to that is to watch the xvid status window during the second pass: if you only see quantizers at 2 and 4 (or 3 depending on b-frames quantizer ratio & offset) that means the codec has 'maxed out' as duartix said, meaning you can't get any better quality for your given resolution and settings

are you certain you hit 'Load Defaults' after installing XviD 1 as koepi said ??? It's really important that you do that!! do that and then set your desired settings...

split710it
8th January 2004, 11:27
Yes Im' sure to pushed LOAD DEFAULT after installation (Installed, open virtualdubmod, compression, Xvid... LOAD DEFAULT, closed virtualdub)

Is this the right way?????

************************************************

Tried again.... problem still open (bug ???? )

Codec: Xvid 1.0 beta 3 (latest one)

Now:

source:
original size: 1619715, VCD (m1v), no audio 320x240 cropped, NTSC
29.97

used: Virtualdubmod 1.5.10 ONLY (no avisynth, no other external prog)

full recompress with NO audio

VIDEO > Frame rate > Inverse telecine (3-2) = Reconstruct from field-adaptive (To inverse from 29.97 t0 23.976)

VIDEO > filters > add > rsize = 544x240 > precise bicubic a=075

VIDEO > filters > cropping (made right crop)

COMPRESSION > xvid codec.... > configure > pushed "LOAD DEFAULT"

I now setted all for two pass (first pass Matrix MPEG, B frame 2, Chroma motion on, VHQ 2, motion search 6, GMC off)

second pass:
SIZE/BITRATE

SIZE desired = 1280000
BITRATE desired = 1270

Finale real job is:

544x240, 23.976 (that's fine)

SIZE = 691
BITRATE = 609 (g.spot)

bad image......

Why still undersize????????

Tried and re-tried with more of one movie always the same............

Could be still available BUG for bif files????????

P.S.
No problem with oldest Koepy codec before version 1.0

Ciao

dayer
8th January 2004, 13:18
Hi, and before all congratulations for this Web and Forum :D

My problem is very similar at split710it, but I'm not interested in SVCD. I do de avisynth proyect loading mpeg2dec3.dll, crop and resized; then load this proyect in VirtualDubMod, I do the codec's ajust, save the two passes but.... I obtain and undersize around 250MB more or less.

I have read others post but I don't find the solution. I have hit in "load defaults" also.

PD: sorry for my poor English :(

sysKin
8th January 2004, 13:24
Frankly I have no idea what's going on.

Could you please (and dayer too) load your statsfile into Koepi's statsreader and tell us what is the size of first pass (it will be there).
Thanks

Koepi
8th January 2004, 13:53
I think setting the target bitrate causes these troubles. don't use bitrate mode. Hit the switch so it shows "target kbyte" and enter 1200000 there.
(NO, don't hit that button again afterwards.)

Maybe we messed something up, so hit Load Defaults again, setup the codec, and don't touch at all that button "desired size"/"bitrate". Just enter the desired size, done.

Regards
Koepi

split710it
8th January 2004, 13:57
Ok Syskin

P.S.
Remember that oldest version do the right final size with same movie..


Problem is now that I cancelled already, I will try again this evening
and I will send you that.


P.S. Strange in any case......

Supposing for mistake a wrong (undersize) starting size (codec see a less size) codec, in second pass, should be INCREASE it to size request in outpou... or not???

But... it is only my suppose.....

I will send this evening it.

Audionut
8th January 2004, 16:02
Originally posted by Koepi
I think setting the target bitrate causes these troubles. don't use bitrate mode. Hit the switch so it shows "target kbyte" and enter 1200000 there.
(NO, don't hit that button again afterwards.)

Maybe we messed something up, so hit Load Defaults again, setup the codec, and don't touch at all that button "desired size"/"bitrate". Just enter the desired size, done.

Regards
Koepi

I have no problems at all encoding using the bitrate mode. Every movie so far for me has been spot on. Actually the average bitrate always ends up about -5kbps under. Pretty much perfect if you ask me.

Also load defaults for me creates these values,

Max overflow improvement% 5
Max overflow degradation% 5
Overflow control strenght% 5

I was under the assumption that these values would impead the codec from being able to hit the desired bitrate?

I used to set them to the old defualts of,

Max overflow improvement% 60
Max overflow degradation% 60
Overflow control strenght% 0

However :)

Doom9 xvid 1.0 guide,

Max overflow improvement and Max overflow degradation indicate how much the codec can compensate during undersized and oversized sections of the video. The higher those values, the faster an overflow / underflow will be corrected (but it also means that no bitrate is preserved for other scenes that might need it more). If you're getting high fluctuations in quantizers (noticeable by fluctuations in image quality), try reducing those values to 20-30. This might hurt size predictability though.

Going by this statement I would assume that setting a slightly lower value of "max overflow" say 30-40 would be better for ensuring that bitrate is preserved for scenes that might need it more, while not hurting size predictability.

Manao
8th January 2004, 17:14
Audionut : the Doom9's guide is a bit outdated, the default value, and the behavior of these figures have changed between beta 2 and 3.

split710it : a second pass size of 610 Mo, for a full movie, in 320x240 is quite normal ( imho ). The bad quality may be due to the quality of the source itself ?

dayer : when you mean undersized around 250 Mb, you mean the filesize was 250 Mb less than expected ? If yes, depending on how big the bitrate was, it may perfectly be the classical undersizing 'issue', when you reach the point beyond which you can't compress less.

split710it
8th January 2004, 17:44
manao wrote

split710it : a second pass size of 610 Mo, for a full movie, in 320x240 is quite normal ( imho ). The bad quality may be due to the quality of the source itself ?

I know that 320x240 for full movie it's enough 610 MB..... this is not the problem.....

I would like to have the size that I desire, not what is enough... (also if is wrong).. that's different.

I'm converting a VCD 320x240 in Xvid 544x240 (knowing that the resulting size cannot be better that original) I can't explanin why.. it's too dificult)

I'm just finding the right way to do it, Now I try in a suggested Koepy way

thank's a lot to all for supporting :)

ciao

mikeX
8th January 2004, 17:50
i don't think xvid is gonna give more size than what it requires with it's current settings (at least with 2 pass encodes, don't know about 1 pass cbr).

you could try putting some extra noise in the source to make the resulting file bigger. there is a post about that somewhere by mf (it also has the required avs script attached to it) but i don't know which one it is, use the search function

split710it
8th January 2004, 18:01
Manao wrote:

dayer : when you mean undersized around 250 Mb, you mean the filesize was 250 Mb less than expected ? If yes, depending on how big the bitrate was, it may perfectly be the classical undersizing 'issue', when you reach the point beyond which you can't compress less.

I give you my experience:

setting final bitrate 1270 resulted final size was as for 607 bitrate
setting final bitrate 976 resulted final size was as for 601 bitrate
setting final bitrate 1045 resulted final size was as for 416 bitrate

first point
bitrate 1270-976-1045 are NOT to big for a XviD compression but normal so not "oversize" there is no reason to create an "undersize" file ?

second point
607-601-416 are not proporthioned ?

1270 : 607 = 976 : x
where X = 466 and not 601 (less bitrate)

1270 : 607 = 1045 : x
where X = 499 and not 416 (bigger bitrate)

that's a second mystere ?

ciao

Manao
8th January 2004, 18:03
split710it : OK, then I understand. The default settings with the XviD beta's are far more effective than with the previous version, concerning the compression of the source. If, with the devapi3, you were able to reach 1200000 Kb, but you were just under the maximal size reachable, then with the devapi4 you'll get undersizing.

What you've said in your posts makes me think you don't know exactly how the compression work in a 2-passes mode. Let me explain quickly. Suring the first pass, the codec encodes the source with the maximum quality it can reach ( with the settings you provided to him ). After the first pass, you can get the maximum filesize reachable with your settings by opening the filestat with StatReader ( we still are waiting for that number ). Without changing the settings of XviD, the second pass size can't be larger than this size.

So if you really want to get a higher filesize ( I don't know why though ), either don't use b-frames, or use MPEG matrix, or even a custom matrix ( Andreas78th for exemple ). But I think it's silly, since the quality of the source isn't great, to try to reach a higher filesize.

@mikeX : that's another way of increasing the filesize, but imho, I think it's a silly way to do it. Adding noise will of course remove blocks that may be present in the source, and it may make it look better, but XviD ( and all MPEG4 codecs ) doesn't like noise. And, you can still add the noise while playing back, which will have the same effects, but without pertubing the codec.

Edited :

split710it : I just saw your post. The first two values are normal ( at least, explainable ), the third one is strange ( I think you made a mistake during your settings on this one ).

split710it
8th January 2004, 18:11
MikeX wrote:

you could try putting some extra noise in the source to make the resulting file bigger. there is a post about that somewhere by mf (it also has the required avs script attached to it) but i don't know which one it is, use the search function

Remember that with older Xvid codec (last year) there was no problem to encode incresing size with every kind of source.. and without to use noise filters. (P.S. It's better to have a undersize file without noise that rightsize with it :) )

My opinion still is that the problem is concentred in the way that Codec use to interpretate the DESIRED FINAL SIZE only and not filters, IVTC, framerate, original size........

ciao

mikeX
8th January 2004, 18:20
haven't bothered myself with noise adding filters, just happened to stamble across that thread sometime ago and thought i'd mention it
btw thanks for the clarifications manao

Manao
8th January 2004, 18:24
Originally posted by split710it
Remember that with older Xvid codec (last year) there was no problem to encode incresing size with every kind of source..Wrong, you just didn't encountered such cases with previous version, but undersizing was possible.

@mikeX : I gave my opinion on using noise before encoding, but some may disagree ( and have concrete basis to disagree ).

split710it
8th January 2004, 18:38
Manao wrote:

What you've said in your posts makes me think you don't know exactly how the compression work in a 2-passes mode. Let me explain quickly. Suring the first pass, the codec encodes the source with the maximum quality it can reach ( with the settings you provided to him ). After the first pass, you can get the maximum filesize reachable with your settings by opening the filestat with StatReader ( we still are waiting for that number ). Without changing the settings of XviD, the second pass size can't be larger than this size.

uhm...... Yes is true I don't know exactly a compression work in 2 pass but....

What you mean with "to edit the stat file" i never done before.. When I compress in two pass I prepare a Job, and after started I cannot stop it..... for first pass I always used values like: 6, VHQ 1 or 2, B frames 2, sometimes chroma ON, no GMC, no alternate curve, no Qpel.....

As you can see it's not the maximum of quality possible but "I never had problem of undersize before using old codec".

reading this your answer you tell that:

If compressing with maximum quality on first pass results a maximum size of 803.000 MB I cannot realize a biggest one in second pass ??

Could be, but let me understund why using "the same datas", the same VCD movie with "same desired final size" with:

XviD-Koepi24062003

final Size is exactly as requred and the quality is REALLY BETTER .
(in that case 1.270.000) ?

What I mean is that size (obtained with XviD-Koepi24062003) is NOT ONLY to big but with REAL BETTER quality (Immediately watchable) of the MAXIMUM UNDER-size obtained with koepy 1.0 beta 3...

So my mind is now in a state of misunderstunfing :)

ciao

split710it
8th January 2004, 18:41
manao wrote:

split710it : I just saw your post. The first two values are normal ( at least, explainable ), the third one is strange ( I think you made a mistake during your settings on this one ).

Yes believe me is true... but i forgot to tell you that in that last case was a different movie. But final Size desired 1287000, resulted size 168000 about, that's really strange.... ?

ciao

Manao
8th January 2004, 18:44
Hum, reread my post, with taking attention to the text inside parenthesis. I said you could not obtained larger filesize without changing the settings. But you still can change the settings, and at the end of my post, I gave the settings which had to be changed in order to achieve this greater filesize.

I also explained why the results were different from previous versions of XviD.

But I still don't think you should try to reach a higher filesize.

Edited : just saw ( again ) your last answer. Then all three numbers are explainable, except the issue with which you came up first ( 1280000->160000 ) where I do think you made a mistake ( or something gone wrong during encode )

split710it
8th January 2004, 19:02
thank's manao for your support

I readed well what you said..... I understood also new things not knew before....

If new codec produce a file that is baddest that the same obtained with a old codec (with ames datats) it would be to think that the newest is BADDEST that the oldest... I don't want to believe and to think that's true so.. I have to realize the solution.

But at the moment I still think that the problem is to interpretate the final size from the Codec only...... At the monet I still think that raeading the stat file I can see that produced final size "could" be my desired size)... I only don't understund why second pass Don't do it..... If is not.... also with oldes codec i must have same problem....

the reason of that is as explaned NEW file (810.000) is BADDEST that file of 1.287.000 done with oldest codec......

I could believe in you at 100% ONLY if the quality is "THE SAME" for the two files.

I try new solution this evening....

ciao

Manao
8th January 2004, 19:14
OK, Give us the size you read with StatReader, without it, it is impossible to find what could have went wrong.

Second, you constantly gives us different filesize( 810000, 160000, 604000 ... ). Please be a little more specific, and try to give each time the first pass size.

For the quality issue, are you sure you speak of the same source, same avs script, same b-frames settings, ...?

mikeX
8th January 2004, 19:39
also, if you want to see the maximum quality the codec can achieve with the settings you choose do this (with xvid 1):
sellect 'Twopass - 1st pass'
sellect your desired settings
click on the second box to the right and make sure 'Discard first pass' is unchecked
click on the zone and select edit
set quantizer to 2
run the first pass (don't put it to job control just run it)

the resulting file will have the best quality the codec can give with your settings (as well as the biggest size)

Manao
8th January 2004, 19:50
@mikeX : even more simple, do an encode at quant 2. Moreover, the first pass is now irrelevant, quality wise ( since beta 3 )

split710it
8th January 2004, 20:09
thank you for all

i'm now at home, I finally understood

difference is in the first pass

I just done now a new rip

original size 1.1163.164 VCD (M1V) 320x240

setted final size 1.200.000

really done 758.000

cheched stat file with stat reader and result maximum file possible with that setting 758 MB.

I understood now..... diference is in the setting of first pass that with a new codec is different of the oldest one, also using similar setting. (probably similar.. but really with some differences)

now I have AVI done with the first pass + sts file, what you suggest me to increase the final size or to start from beginning???
(i would likt to increase if quality will be better)

I have a second problem.. video is interlaced i also have dto deinterlace, could i do from first pass to second?

ciao

Manao
8th January 2004, 20:16
It's useless to increase the final size. If you really want to get a larger first pass size ( and so being able to get a higher finalsize ), don't use b-frames.

You have to do the whole process again to deinterlace ( else the first pass made without deinterlace would be useless ).

Without notice quality drop, at equivalent settings, you can safely lower the second pass size to 80% of the first pass size ( in your case, you could aim at 758*0.8 = 462 MB ).

With such a difference between previous codecs and this one, you must have changed some settings ( I would say b-frames, vhq 4, and perhaps a different matrix ).

Why did you want to transcode ( converting from MPEG to XviD ) the video if reducing size isn't your goal ?

mikeX
8th January 2004, 21:13
even more simple, do an encode at quant 2.

that's what i was talking about (zone @ quantizer 2)

Moreover, the first pass is now irrelevant, quality wise ( since beta 3 )

i believe sysKin explained to Leak somewhere in the 'Selam' thread that it would be possible to make 1st pass act similar to the previous betas by doing exactly as i described (zone @ quant 2)

what i would really like to know is if there is some difference in quality between 'Single Pass' @ quant 2 and 'Twopass 1st pass' @ quant 2 (taking the 'reaction delay factor' etc settings for 'Single Pass' into consideration)

Manao
8th January 2004, 21:37
MikeX, you're right, I didn't see the part in your post when you said to create a zone at fixed quant 2. In that case, the output should be the same.

digitalman
9th January 2004, 04:04
As I stated in my earlier post I am having the exact same issues as split710it with Avisynth. But when I load the exact same M2V directly in VirtualDubMod, I do not have the undersize issue. It comes out at the size I set, 1800000KB. When I load it using Avisynth and MPEG2DEC3.DLL, the movies come out in size between 12000000 and 1500000KBs not 1800000KB. I will try Koepi's suggestion on how to fix it.

sysKin
9th January 2004, 04:56
Originally posted by mikeX
what i would really like to know is if there is some difference in quality between 'Single Pass' @ quant 2 and 'Twopass 1st pass' @ quant 2 No, the only difference is that stats file is created.
:)
Radek

Didée
9th January 2004, 09:18
Originally posted by sysKin Originally posted by mikeX
what i would really like to know is if there is some difference in quality between 'Single Pass' @ quant 2 and 'Twopass 1st pass' @ quant 2 No, the only difference is that stats file is created.
:)
Radek
Syskin, how does your reply fit to your sig's FAQ #1 ?

- Didée

sysKin
9th January 2004, 10:11
Originally posted by Didée
Syskin, how does your reply fit to your sig's FAQ #1 ?FAQ just doesn't contain this trick, becuase it's a lousy trick, which takes unfair advantage of quant zones :p

dayer
9th January 2004, 12:20
@sysKin:
Sorry for reply later.
Desired size: 1218497 KB
Final size: 954946 KB
Size in stat: 928 MB
(all without audio)

And the .avs is:

#########################################################
# Script Generado por AR Calculator #
# #
# http://www.canalxvid.com #
#########################################################

LoadPlugin("D:\rip\ARCALC~1\mpeg2dec3.dll")
LoadPlugin("D:\rip\ARCALC~1\undot.dll")
LoadPlugin("D:\rip\ARCALC~1\Convolution3D.dll")

Mpeg2Source("D:\DOCUME~1\dayer\ESCRIT~1\rip\PANORA~1\ppm.d2v")

LumaFilter()
Crop(0,70,-0,-76,align=true)
Undot()
Convolution3d(preset="movieHQ")
BicubicResize(576,240,0,0.5)
------------------------------------------------------------------

Could be this problem for use AR Calculator?

PD: sorry for my poor English :(

Mug Funky
9th January 2004, 12:42
@ dayer:

that undersizing seems to me like simple codec saturation.

try putting a # in front of the line that says "convolution32". that'll make the file bigger :) you'll lose less detail too.

as for the discrepancy between the final size and the stats size... well 954946/1024 = 932.6MB, which isn't far off the stats size at all (the rest is probably avi frame overhead). i don't see much of a problem with it anyway. i figure if a file is undersized at quant 2, the quality will be transparent anyway. why worry about megs? if you can't tell the difference there's no probs, and it's not like CDs are expensive... just put something else in the remaining space if you're worried about wasting the rest of the CD (which would be wasted anyway by bloating the Xvid file).

@ split710it:

i don't have enough detail to understand your problem yet, but it looks like you've solved it :)

dayer
9th January 2004, 12:52
Thank's very much Mug Funky, I have used some filters for view if the size change too much. I think that the best is use the minium filters possible (mpeg2dec3.dll, crop and resized). The megs is because or use 1CD or 2 CDs, but 1 CD and half of other... better use an only CD.

Congratulations for the speed in reply in this forum :D

split710it
9th January 2004, 13:15
I tried yesterday evening some things..

Following is what I understood encoding in two pass

First pass determinate in any case the maximum size possible (with done setting offcourse), with no rilevant changes, second pass it could only do SAME or LESS size but NEVER to big.

Starting with this idea I realized two solution for big input size files:

1 - There is no reason to look for a oversize in second pass If the first pass setting are enought for yyy movie (also if not setted for absolute maximum quality).

2 - I can set first pass with maximum quality possible to try to have a size very similar to desired in second pass.

Now question is.....

what the two best solution to have:

larger size possible (if i don't have absolute limited size problem)
right size (if I have limited size problem)

I tink this:

I can set MY DESIRED size in a Size Tab space (second pass) If stat answer (after first pass using maximum quality possible)
give me a outside size just SAME or little BIG that desired without changing setting

If answer of stat give me a VERY BIG size realizable what you suggest?? to decrease setting for second pass (setting desired size on space also) or just to set only desired size....

If using maximum quality answer of stat give me undersize did you think is a good idea to reencode increasing resolution ????

just to finish.... i read a doom9 document about setting of codec 1.0 beta 1... but... suggest me in really two words what the better setting to have real better quality in first pass

sorry for my english :)

ciao

Manao
9th January 2004, 17:18
Another thing you seem not to know about two passes algorithm : you must not change the settings between first pass and second pass. By settings, I mean almost every setting of the codec ( except the settings specific to first and second pass, mainly ), and the settings of avisynth ( resizing, resolution, filtering )

So, if your first pass is too small, and you want absolutely a bigger filesize, you have to make again a first pass and a second pass with different settings.

If the first pass size is 'too big' ( the definition of too big depends on the person : I consider that if the first pass size is twice as big as the second pass size, it is too big. Less than 2 is OK ), you also have to make again first and second pass.

Now, if you have undersizing, is it interesting to increase the resolution ? I don't know, because it depends of what you want to do. If size really doesn't matter, yes increase it slightly. But instead, you could for example try to put the movie on a single CD, instead of 2.

You want better quality in first pass ? The following will increase quality, if size doesn't matter :
- Disable B-frames
- Change the matrix of quantization ( do a search to see what it means, and try a matrix like Andreas78, or MPEG if it was not already the one you were using )

split710it
10th January 2004, 02:26
thanks a lot again :)
manao wrote:

Another thing you seem not to know about two passes algorithm : you must not change the settings between first pass and second pass. By settings, I mean almost every setting of the codec ( except the settings specific to first and second pass, mainly ), and the settings of avisynth ( resizing, resolution, filtering )

What I mean is not to change setting but to set what is possible to do in the second pass, for example the desired size. I think codec do the best to eventtually reduce from maximum quality to enought quality for side vrote on the Tab..... There is no reason to do two pass if is not !
Are you also writing I can change resolution between first and second pass??? I was not sure at 100% :)... only avisynth filter or virtualdub filter too (I mean about resolution, resize)

You want better quality in first pass ? The following will increase quality, if size doesn't matter :
- Disable B-frames
- Change the matrix of quantization ( do a search to see what it means, and try a matrix like Andreas78, or MPEG if it was not already the one you were using )

I normally try to have maximum quality considering that it's a Xvid file so I consider everything (quality-size-cd used, etcc) Sometimes I want to have a maximum quality so I will encode in 2 CD instead 1 also for a movie nor very big, like 1 hour, 30 minutes. Normally not.

until 1 hour, 45 min I use 1 CD
more of that depends, I use for sure 2 CD for more of 2 hours (like movie that used starting this thread 02:37:40:021.

I normally use B-frame 2 always, disabled for 2 CD with not big time movie. I normally use MPEG matrix, I don't use Qpel (not compatible with my Kiss DVD player)

Thank's a lot
Ciao

Manao
10th January 2004, 09:56
Are you also writing I can change resolution between first and second pass???No, I said the contrary ( but that wasn't very clear, I admit it ). I said not to change the settings, and then I defined the settings by almost every setting of the codec, and by avisynth filters. So you must not change avisynth filters between first and second pass.

sysKin
11th January 2004, 05:46
Originally posted by Manao
So, if your first pass is too small, and you want absolutely a bigger filesize, you have to make again a first pass and a second pass with different settings.Ever since beta 3, you can also allow quant 1 to be used - which will make your second pass bigger than first. Just set max overflow improvement/degradation to 20% (or more) or else overflow will never be strong enough to control anything.

Increasing resolution will give you better results though.

I'll investigate the undersizes you all are experiencing... I've always thought it's about max overflow effect, but I might have been wrong.

Radek

ang3l
14th January 2004, 14:55
also I have a large problem of undersize with the new xvid 1.0 (all beta rls)

this is the script of avisynth:

LoadPlugin("C:\PROGRAß1\UTILITß1\GORDIAß1\mpeg2dec3.dll")
mpeg2source("C:\test\test.d2v")
crop(0,74,720,428)
LanczosResize(640,272)

I set the desidered size to 631753,

qpel activated
disabled the b-frame
vhq-off
no trellis
quantization type MPEG
Quantization offset 0.75
max overflow improvment / degration 60
overflow controll 10

turbo mode ;-) activated

(I push load default before to set the codec)

but my final result is always around the 500MB

I try also other film and almost always to the end of the first pass the statsreader marks like greatest size 500 / 450 mb...

Why it happens?

I cannot believe that almost all of the film it go in saturation with this codec with this settings

thx and excuse for my english

stax76
14th January 2004, 17:01
I think setting the target bitrate causes these troubles. don't use bitrate mode

what's wrong with bitrate mode? Also I would like to suggest a additional MB mode and to remove the AVI overhead correction. AVI is not the only container on this planet, I reportet this several times now. It's definitively not the job of a codec to take care of the container overhead

Koepi
14th January 2004, 19:59
Originally posted by Dolemite
what's wrong with bitrate mode? Also I would like to suggest a additional MB mode and to remove the AVI overhead correction. AVI is not the only container on this planet, I reportet this several times now. It's definitively not the job of a codec to take care of the container overhead

Nothing wrong with bitrate mode, but I've often seen that people have troubles with the necessary calculations. Call it human processing error if you like ;)

MB mode is very simple: take your desired MBs and multiply with 1024. It's very simple and I don't see where it differs at all. Maybe I'm too much of a mathematician or computer scientist.

To hit the desired size it is necessary to include container overhead. You know, 99.999% of the people here encode to avi and mux the files into different containers later on. They would be _very_ annoyed if their files were always oversized by several MBs. (I think I answered this to every "report" of yours now, didn't I?)

Regards
Koepi