logo Sign In

GOUT image stabilization - Released — Page 17

Author
Time
 (Edited)
iRantanplan said:

A week of encoding, omg. What kind of maschine do you have?

 

 dual core 2, 1.8 GHZ, with 2 gigs of RAM.... (windows XP)..

but i was also running a ton of other stuff at the same time,

like converting mkv->avi, virtualdub scripts, bittorrent, etc, etc..

i didn't want to use up all my processer time with just this, it put it at 50% normal priority.

i'm sure i could have sped it up, but i still needed to do other stuff on this computer..

 

here's a partial log of the encode of the 1st of 7 VOB files:

----------------


              -----------------------------------------
              | HCenc - MPEG2 encoder - rel. 0.23.0.0 |
              -----------------------------------------

MPEG profile@level:    MP@ML
input:                 o:\data\movie\dvd\star wars ep 4-gout\gout309.avs
output:                C:\data\vid\movies\gout\gout.m2v

              --------------------
              | encoder settings |
              --------------------

profile:               NORMAL
frames:                0 - 34834
framerate:             29.970
aspect ratio:          4:3
bitrate Kb/s:          2000
max. bitrate Kb/s:     8000
pulldown:              no
closed gops:           no
VBV check:             yes
scene change det.:     yes
interlaced:            auto, TFF
goplen,B-pic:          AUTO
dc_precision:          9
scan method:           auto
bias:                  0
chapter frames:        0
time code:             0  0  0  0
CPU:                   SSSE3
priority:              idle
SMP active:            yes
matrix:                MPEG
luminance gain:        no
adaptive quantization: yes, strength 1

              ------------------
              |  source stats  |
              ------------------

nr. of frames in source:         34835
width*height:                    720x480
fps:                             29.970
nr. of frames to encode:         34835
frames to encode:                0 - 34834

              ---------------------
              | encoding - pass 1 |
              ---------------------

pass 1 encoding time:            8:18:52 (29932.19 s)
fps:                             1.2

              ---------------------
              | encoding - pass 2 |
              ---------------------

pass 2 encoding time:            8:17:00 (29819.90 s)
fps:                             1.2

              ------------------
              | encoding stats |
              ------------------


nr. of gops:          2508
nr. of frames:        34835
nr. of I-frames:      2508
nr. of P-frames:      9641
nr. of B-frames:      22686
average bitrate:      2000
minimum bitrate:      148
maximum bitrate:      7999

bytes in bitstream:   290751811
bits in bitstream:    2326014488

average Quantizer:    6.913

total CPU time:       18:10:21 (65421.08 s)

later

-1

[no GOUT in CED?-> GOUT CED]

Author
Time
negative1 said:

yes, i had to use some deinterlacing here and there in the final version, but it still looks good to me

 hmm, that doesn't sound right... I don't understand what would need deinterlacing.

negative1 said:

now i'll work on upscaling to 720p and 1080p with the solo21 script, and some more modifications..

hmm, that would add a lot of "lossy" and redundant steps. That's a lot of resizes and potential motion-compensation errors for something that that already doesn't even have enough detail for DVD quality.

-G

 

Author
Time
g-force said:
negative1 said:

yes, i had to use some deinterlacing here and there in the final version, but it still looks good to me

 hmm, that doesn't sound right... I don't understand what would need deinterlacing.

negative1 said:

now i'll work on upscaling to 720p and 1080p with the solo21 script, and some more modifications..

hmm, that would add a lot of "lossy" and redundant steps. That's a lot of resizes and potential motion-compensation errors for something that that already doesn't even have enough detail for DVD quality.

-G

 

 

about the interlacing, it probably came from the original VOB's?

 

and about the resizing, you're probably right..

it's better to combine your script, and the resizing all in one step, but i'm not sure

how much of a difference it will make in the end by doing it separately..

 

well anyways, i'm still working on it..

i have found a pretty good uprez script that i will tweak

(called SD2HD, and HD2SD).. mentioned here:

---------------------------------------------------------

http://forum.doom9.org/showthread.php?p=1210858#post1210858

later

-1

 

 

 

[no GOUT in CED?-> GOUT CED]

Author
Time

Finally got around to trying this, and wanted to second the interlacing issue.  Sorry if this has been addressed before, but I didn't see an answer as I went through the thread.  Or rather, I'm not sure if what I'm seeing is just normal 3:2 pulldown, and not to worry about.  When I run the script out of Virtualdub, I'm definetly seeing interlacing for two frames, after three progressive.  I've been trying to run a lossless AVI as per g-force and Moth3r's advice.  When I check screenshots of a given frame between the demuxed video, and the lossless AVI, or the mp2 that CCE spit out, there is again definetly interlacing on frames in the AVI where it isn't in the original video.  Is there some setting I have missed in ripping the DVD, or is this nothing out of the ordinary?

Author
Time

There should be no interlaced (or combed) frames when previewing the script with VirtualDub, or in any intermediate lossless AVI you create. The framerate should be 23.976fps, and every frame should be progressive. I suspect you didn't use the "Force film" option in DGIndex when creating the d2v file.

To be sure it's your source and not the script, try a script with just the line:

Mpeg2Source("\path\filename.d2v")

and check for combed frames.

The final MPEG-2 encode for DVD will end up being 29.97fps with 3:2 pulldown. Depending on how you view/open this video, you may see some combed frames in this file.

Guidelines for post content and general behaviour: read announcement here

Max. allowable image sizes in signatures: reminder here

Author
Time

Argh...'Tis exactly what happened.  :)  I was just coming on to edit my post.  I was following the directions in the Gout to Anamorphic thread, which doesn't seem to mention the "Force Film" option in the indexing stage.  Looks fine after I re-indexed it.  Only problem now is that VirtualDub can't seem to make it all the way through outputting as lossless.  The script ran for about 10 hours, and got as far as looking for R2, and came up with an AVISYnth error, which seems to be a memory problem ("could not reference...").  I'm trying a couple of tests now straight out of HC, checking video and also to see if subs appear, which they didn't seem to in Vdub, hang the extra time it'll take.  Thank, Moth3R!

Author
Time
 (Edited)

yeah, thanks mother,

 

that seemed to be the issue with mine also.

looks like i will have to do a reencode..

 

when you mentioned framerate : mine came

out at 23.976024  , its the NTSC widescreen version,

i want to make sure its right... you didn't round it

off  to 23.976 did you???

 

later

-1

[no GOUT in CED?-> GOUT CED]

Author
Time
INv8r_ZIM said:

Argh...'Tis exactly what happened.  :)  I was just coming on to edit my post.  I was following the directions in the Gout to Anamorphic thread, which doesn't seem to mention the "Force Film" option in the indexing stage.  Looks fine after I re-indexed it.  Only problem now is that VirtualDub can't seem to make it all the way through outputting as lossless.  The script ran for about 10 hours, and got as far as looking for R2, and came up with an AVISYnth error, which seems to be a memory problem ("could not reference...").  I'm trying a couple of tests now straight out of HC, checking video and also to see if subs appear, which they didn't seem to in Vdub, hang the extra time it'll take.  Thank, Moth3R!

 

 have you tried encoding each VOB separately?

that way you don't have to do the whole thing at once,

and maybe you won't have the memory issue..

 

later

-1

[no GOUT in CED?-> GOUT CED]

Author
Time
negative1 said:

have you tried encoding each VOB separately?


Sounds like a recipe for dropping a frame or three then scratching one's head wondering why the audio went out of sync.

If you try it, make sure each VOB begins with a navigation pack.  Many times a GOP can span two VOB files.

However, in practice you must take into account the “fuckwit factor”. Just talk to Darth Mallwalker…
-Moth3r

Author
Time
 (Edited)
Darth Mallwalker said:
negative1 said:

have you tried encoding each VOB separately?


Sounds like a recipe for dropping a frame or three then scratching one's head wondering why the audio went out of sync.

If you try it, make sure each VOB begins with a navigation pack.  Many times a GOP can span two VOB files.

i've done it, and never had a problem (yet)...

the new version of DGMPdec seems to take that into account.

 

and also, when using HCencoder, it shows the frame numbers,

so you can match those also to make sure its there..

 

i have not had a problem with the audio encoding, as i don't

think 3 or 4 frames would throw it off that much  (i think) even if

 it did happen ...

 

later

-1

 

[no GOUT in CED?-> GOUT CED]

Author
Time
 (Edited)
INv8r_ZIM said:

Argh...'Tis exactly what happened.  :)  I was just coming on to edit my post.  I was following the directions in the Gout to Anamorphic thread, which doesn't seem to mention the "Force Film" option in the indexing stage.  Looks fine after I re-indexed it. 

Thought so. If you look back through this thread I think there were a couple of other posters who had the same problem.

@Oldschooljedi:

Only problem now is that VirtualDub can't seem to make it all the way through outputting as lossless.  The script ran for about 10 hours, and got as far as looking for R2, and came up with an AVISYnth error, which seems to be a memory problem ("could not reference...").  I'm trying a couple of tests now straight out of HC, checking video and also to see if subs appear, which they didn't seem to in Vdub, hang the extra time it'll take.  Thank, Moth3R!

Do you have any "SSE3" DLLs (RemoveGrainSSE3.dll, RepairSSE3.dll) in your AviSynth plugins folder?

Guidelines for post content and general behaviour: read announcement here

Max. allowable image sizes in signatures: reminder here

Author
Time
 (Edited)

@Oldschooljedi:

 

I read earlier in this threat, that "Force Film" should be selected. But I was a littlebit confused about that as I'm using the PAL-DVD's.

When I run DGIndex with the PAL-DVD I get this message:

"It is almost wrong to run Force Film on an MPEG 1 stream, or a stream with a framerate other than 29.970. Your stream is MPEG 2 and has a frame rate of 25.000. Do you want to do it anyway?"

So I used DGIndex without changing anything there and everything worked fine.

This was the reason, I decided not to add this step to this guide. Now I've learned that this is very impossible for encoding the NTSC-DVD's and I will add this step into this guide.

 

Author
Time

Why...yes I do, as matter of fact (have both dll's you mentioned, Moth3R).  I know I've picked up a couple of packs specifically for the GOUT re-encode along the way.  I do remember seeing that a couple of variations on the plugins required were causing trouble for other people, were these the offending DLL's, those need to go?  It just seems weird that the encode made it halfway through before quitting.

Author
Time

I've got some updates to the script that I have been holding off posting until they were tested. Testing turned out great. They include better anti-aliasing, and some improvements that greatly help the last issue that OSJ had. As usual, the updated script is on the first page of this thread.

-G

Author
Time

I just watched the cantina-scene encoded with the actual script, this looks really superior compared to the encodings with the older script. Good work, g-force.

Author
Time
INv8r_ZIM said:

Why...yes I do, as matter of fact (have both dll's you mentioned, Moth3R).  I know I've picked up a couple of packs specifically for the GOUT re-encode along the way.  I do remember seeing that a couple of variations on the plugins required were causing trouble for other people, were these the offending DLL's, those need to go?  It just seems weird that the encode made it halfway through before quitting.

At one point there was a bug in the Removegrain SSE3 module that caused crashes, although I believe it has now been fixed. Make sure you have the fixed version. (I assume of course you have a CPU with SSE3?)

I agree it's strange that it encoded for several hours before crashing. You could try encoding again, starting at the problematic scene, to see if it's something in that part of the video that caused the crash.

It could also be a stability or overheating problem. Have you carried out any stress-testing of you system using, for example, Prime95?

Guidelines for post content and general behaviour: read announcement here

Max. allowable image sizes in signatures: reminder here

Author
Time

I'd do it myself, but I'm booked up for work for weeks, but anyway, just to throw in the idea - As for motion smear, there are a lot of unchanged shots that could be taken directly of the 2004 SE DVDs. How about writing a script to color-correct and replace most of those shots? Anybody who has the GOUT on Disc 2 has the SE on Disc 1 as well. That could easily be a team effort directly here on the board.

Author
Time
resonator said:

I'd do it myself, but I'm booked up for work for weeks, but anyway, just to throw in the idea - As for motion smear, there are a lot of unchanged shots that could be taken directly of the 2004 SE DVDs. How about writing a script to color-correct and replace most of those shots? Anybody who has the GOUT on Disc 2 has the SE on Disc 1 as well. That could easily be a team effort directly here on the board.

I think that's a great idea, and is actually something I've been thinking about for some time. For me, it's going to have to wait until I get a faster PC and larger hard drive, but someday...

-G

 

Author
Time
resonator said:

I'd do it myself, but I'm booked up for work for weeks, but anyway, just to throw in the idea - As for motion smear, there are a lot of unchanged shots that could be taken directly of the 2004 SE DVDs. How about writing a script to color-correct and replace most of those shots? Anybody who has the GOUT on Disc 2 has the SE on Disc 1 as well. That could easily be a team effort directly here on the board.

 

i think there would problems in matching up the shots...there are a lot of them.

i started to try marking all the scene changes in episode 4, but there so many...

 

also the color correction is very difficult, so i would have a hard time seeing

a combined effort that didn't look very smooth in its transitions etc..

 

not saying it can't be done, but it would be very very hard, i think..

 

later

-1

[no GOUT in CED?-> GOUT CED]

Author
Time

I think it might be a stability issue, actually, Moth3r; still different points at which the operation crashes.  Would the script stop automatically if a dll was missing?  Also, is it possible to append chunks of lagarith encoded avi together without frame-drops?  I'm still running into problems with the enocde, at least paritally now becuase I live in a fairly remote area, and power surges pretty commonly knock out my computer.  If I can at least salvage the past day worth of processing, that would be cool.  Any thoughts?

Author
Time

Did you remove those SSE3 plugins?

How much memory do you have? If it's 1GB or less, you could try inserting the line:
SetMemoryMax(256)
at the very start of the script.

Yes it's possible to join lagarith-encoded AVIs with frame accuracy. You can encode smaller chunks either by setting start/end points in VirtualDub, or adding a Trim() command at the end of the script.

However, if your PC happens to crash before VirtualDub has finished saving out that chunk, then that AVI will require an index repair to be useable.

Guidelines for post content and general behaviour: read announcement here

Max. allowable image sizes in signatures: reminder here

Author
Time
INv8r_ZIM said:

I think it might be a stability issue, actually, Moth3r; still different points at which the operation crashes.  Would the script stop automatically if a dll was missing?  Also, is it possible to append chunks of lagarith encoded avi together without frame-drops?  I'm still running into problems with the enocde, at least paritally now becuase I live in a fairly remote area, and power surges pretty commonly knock out my computer.  If I can at least salvage the past day worth of processing, that would be cool.  Any thoughts?

Try this:

Remove all the plugins in you Avisynth-folder and replace them with the .dll's from here.

Then insert the DGDecode.dll from your DGMPGDec-folder.

Maybe this helps.

 

Author
Time
 (Edited)

Tried the script today. Thanks g-force, good job! Very power hungry on my AM2 based-desktop, gotta try it on my Macbook Pro, see if Core2Duo has a little more power.

The THX edge enhancement halos are still noticeable, ranging from barely to terribly. It seems as if no matter what video source you use, there's always SOME thing really wrong with it. So far I've seen Pre-THX LDs which have very little detail, THX-LDs with DNR-smear and the GOUT... no matter how much effort

If we ever want to see the original Star Wars in HD (and with small TV starting at 32" and our viewing habits getting used to more and more HD we'll NEED it in HD soon enough) we'll need something like a cleaned up and color-corrected telecine of a theatrical print.

Kinda frustrating.

 

ONE QUESTION THOUGH - If I intend to upscale to let's say 720p with InstantHD in AfterEffects, should I feed it the anti-aliased NEDI-doubled-in-size video or with the anti-aliasing commented out?

 

Author
Time
resonator said:

The THX edge enhancement halos are still noticeable, ranging from barely to terribly. It seems as if no matter what video source you use, there's always SOME thing really wrong with it. So far I've seen Pre-THX LDs which have very little detail, THX-LDs with DNR-smear and the GOUT... no matter how much effort

The GOUT DVDs were unfortunately from the same source as the NTSC THX LDs. Not only is that awful smearing retained, they also added new flaws such as the aliasing and vertical blur.

The pre-THX LDs do not lack that much detail, but the image does exhibit more dirt and a "greenish" colouring.

If we ever want to see the original Star Wars in HD (and with small TV starting at 32" and our viewing habits getting used to more and more HD we'll NEED it in HD soon enough) we'll need something like a cleaned up and color-corrected telecine of a theatrical print.

I'm willing to accept some of the minor changes present in the 2004 version (for which a HD source exists) - e.g. the digital composition, the tractor beam controls, the cell block corridor - just to have something in HD. Of course it all needs colour correcting, but Adywan has done that already. 

The new CGI shots and Han/Greedo exchange would have to be replaced with video upscaled from GOUT or LD source (maybe a hybrid of the higher-resolution PAL LDs, using the pre-THX LDs only where necessary to avoid the DVNR smearing).

ONE QUESTION THOUGH - If I intend to upscale to let's say 720p with InstantHD in AfterEffects, should I feed it the anti-aliased NEDI-doubled-in-size video or with the anti-aliasing commented out?

You'll have to compare the two methods to see which works best, but I'm guessing that leaving in the EEDI2/NNEDI line and removing the resizing (downscaling) section that follows it would give the best source to feed into AfterEffects.

Guidelines for post content and general behaviour: read announcement here

Max. allowable image sizes in signatures: reminder here

Author
Time

Bwahahahahaha!!!  It's so pretty!  Thanks, guys, I eneded up using the partially encoded first half (no index repair, as nothing would either a)load it or b)found errors, but, aside from the obious pre-load slowness, no problem in virtualdub) appended to a new second half, then full-processing re-encode, staying in lagarith.  I just got the results out of CCE, and it's so much slicker than the source.  Great looking job, g_force, thanks for your work on the script!