AnimeIVTC() - All-in-one solution for interlacing and blending - v2.00 [2010-01-07]

[thetoof]

I came back temporarily last night, heading out again this afternoon and I should be back for good around next week. The bug corrections, new features implementation, discussion with neuron2 and support with the script will resume in max 2 weeks.
I read most of the posts and I've got a solution for everbody + a few other ideas, so see you all in a short while!
Sorry for the trouble.

[Nightshiver]

Great!
Edit: Nvm

[Guest]

I used it on a source where someone handed me a clip...also... heres a before and after of the clip I had to fix (mind you the original source isn't this bad... this mess is the result of someone else in their attempts of handling the source) We don't allow such discussion here, so I have deleted your post. Please read and follow our forum rules.

[Hadien]

We don't allow such discussion here, so I have deleted your post. Please read and follow our forum rules.

wait, just reread the rules and I'm not sure which rule I broke...

are you saying the post was off topic?

the post was basically just me praising thetoof for such a wonderful filter and that it was effective at deblending frames/fields in even an extreme example.

[Nightshiver]

where someone handed me a clip

Look at the bold words....

EDIT: Also, thetoof, some other suggestions.
Update the script so that its compatible with tempgauss_beta1
And possibly update the guide about the beginning parts that talk about "dgindex tell you blah% film" because for me, the latest dgindex (1.5) doesn't tell that anymore.

[RsVolcom]

I'm trying to follow the instructions for mode=5 (hybrid):
Hybrid (mode=5 or 6)
...
pass : Processing a hybrid clip requires two passes.

* 1 : Open with vdub, file-run video analysis pass, wait until the end, close vdub. (default)
* 2 : Reload the same script with pass=2 and do a lossless rendering pass. Encode. Then, take the "timecode.txt" file and add it as the timecode for the video stream in mkvmerge.

So how do I make a "lossless rendering pass"?
Using VDub or MeGui?

Actually once I change 'pass=1' to 'pass=2' I can no longer open the avs in VDub nor MeGui. I get this error:AviSynth open failure
TDecimate: tfmIn error(mode 5, all frames must have an entry)!

[Nightshiver]

Well, in MeGUI, there's an option to do an "Analysis Pass" and also a checkbox for "pre-rendering pass". I believe you would want to do the Vdub method though. I'm using Vdubmod, and there's no "run video analysis pass" under the File menu.

[Adub]

Yes, use the latest virtualdub. Completely drop virtualdubmod, as it is old and outdated and a half-hack anyways.

[HymnToLife]

I'm trying to follow the instructions for mode=5 (hybrid):

So how do I make a "lossless rendering pass"?
Using VDub or MeGui?

Using VDub, just encode your file as you would normally do, either as Uncompressed or compressed with a lossless codec like Lagarith.

[RsVolcom]

I upgraded from Vdub 1.8.1 to 1.8.4, but I get the same error message when trying to load the "pass=2" avisynth script.

I installed the Lagarith codec and I can choose it in VDub, but still get the same error.

[Adub]

The first pass is just an analysis pass for TIVTC.

I don't know what HymnToLife was suggesting. But just do this:

With the Pass=1 script, just drag and drop onto Virtualdub. Under "File", click "Run Video Analysis Pass". What this does is runs through the entire video, each frame, which is then logged by TIVTC to another file.

Then with the Pass=2 script, just load that into what ever your favorite encoding program, whether that is Virtualdub, MeGUI, or something else.

Note: the same thing can be done in MeGUI by pressing the "Queue analysis pass" button on the main page with the Pass=1 script loaded. Once you have run that, change the script to pass=2, and run normally.

[HymnToLife]

The first pass is just an analysis pass for TIVTC.

I don't know what HymnToLife was suggesting. But just do this:

With the Pass=1 script, just drag and drop onto Virtualdub. Under "File", click "Run Video Analysis Pass". What this does is runs through the entire video, each frame, which is then logged by TIVTC to another file.

Then with the Pass=2 script, just load that into what ever your favorite encoding program, whether that is Virtualdub, MeGUI, or something else.

Note: the same thing can be done in MeGUI by pressing the "Queue analysis pass" button on the main page with the Pass=1 script loaded. Once you have run that, change the script to pass=2, and run normally.

Yes, what I was saying is that pass=2 should be encoded normally. Usually, it is encoded losslessly because further processing will be performed afterwards, but if you don't need any, you can encode it using whatever coded you want.

[RsVolcom]

Dang it I tried doing the analysis pass in MeGui but when I switch to pass=2 and try to queue it up I get the same error still.

[HymnToLife]

Dang it I tried doing the analysis pass in MeGui but when I switch to pass=2 and try to queue it up I get the same error still.

Most likely a bug in AnimeIVTC(). Can't say for sure though, I never needed that mode so far. I guess you'll have to wait for thetoof ;)

[RsVolcom]

I guess this is a question for thetoof as well, but maybe somebody will know...what if I have a pattern like 4-4, 4-1, or 5-3, but dgindex tells me Film 75%?

This doesn't seem to coincide with any of the "possible situations" listed in the AnimeIVTC documentation. I want to pick "case 2" (no pattern) but that one is restricted by "close to 100% film". All the cases dealing with less than 95% require a 3-2 pattern, which this is definitely not.

[thetoof]

I'll need a sample to run some tests... and for the rest of you, I'm back in business and I'll be working on animeivtc tomorrow.

[RsVolcom]

OMG I'm so happy to see you're back! Thanks again for this awesome filter...when it works it is by far the best method I have found for encoding these Simpsons DVDs.

I will upload a sample tonight when I get home from school.

Thanks again!

[egrimisu]

Hi the toof i have this anime dvd that has 30p frames then 30i and then again 30p. I read the manual and the script setting that i used is :
animeivtc(mode=6,omode=2,pureint=4,estart1=0,eend1=31287,istart1=31288,iend1=33096,estart2=33097,eend2=34575,aa=0) ; search=3 shall be used only for mode 3??? (3:2 pattern)
if i use omode=3 i get = video frame rate doesn't match
if i use omode=2 i get = there no function named "isodd"
Omode=1 is doing kind a good job but the output is 23.97fps and i can see the motion is not that fluid as frames are missing( is this because i used mode=3 instead of mode=6 and frames were dropped because of that(i belive mode=3 telecined the source).

I'm not sure what omode to use 2 or 3(this is after fixing the issod and framerate errors).

Please take a look at the sample and corect me if i'm wrong somewhere. How to use the 2 pass encode to get the timecode txt (please be very specific as i'm a dumbass and i do not understand very good)

http://www.megaupload.com/?d=78OO9H22

Offtopic( since a source of mine was 99film i used force film but after encodig them i saw a lot of interlancing frames probably wrong trasfer i realy don't know how this can heapen these days. BUT i used honnor flags in dgdindex and i used animeivtc mode=1 AND all the errors disapeard :D THANKS THETOOF FOR THIS GREAT FILTER... the only interlancing frames that remained were is in the source where 3 interlancing frames 1 after another, but those wewe wery rare , 1 maximum 2 in an episode(affordable)


Thanks in advance.

[Nightshiver]

It says it in the guide. You would want to use omode=2 if you want the framerate to be 29.97, and use omode=3 if you want 59.94

[egrimisu]

for now none is working as i get the errors i writen. Since the source has 30i i what i need is 59.94 ? (bobbed)

[Nightshiver]

It all depends on what you want. You can go either way. 59.94 will give you the full framrate.

[thetoof]

The new version will have to wait a bit cuz I had some major bug and I have to happily spend time recovering data and reformatting my comp...

@ egrimisu
Progressive = no need to process
Interlaced = Deinterlace
Nothing telecined here, so you don't need IVTC and it can easily be done externally. I don't see the point of adding a mode for this in AnimeIVTC because it's only about deinterlacing a part of a clip.
For vfr encoding with full temporal resolution
s=mpeg2source("whatever.d2v")
a=s.trim(0,31287).changefps("ntsc_double")
#you need to double the framerate by duplicating frames to have the same framerate for the whole clip. They will be removed anyways with vfr decimation

#uncomment one of them (test and pick the best looking one)
#b=s.mcbob().trim(62576,66192)
#b=s.tempgaussmc_beta2(edimode="nnedi").trim(62576,66192)

c=s.trim(33097,34575).changefps("ntsc_double")
a+b+c
#run this script fully as an analysis pass, remove line for pass1 and uncomment line from pass2, load to encode
#basic parameters for tdecimate, add more if you feel like it
#pass1
TDecimate(mode=4, output="stats.txt",denoise=true)
#pass2
#TDecimate(mode=5, hybrid=2, input="stats.txt", mkvout="timecodes.txt",denoise=true)
For cfr encoding @ 29.97 fps (or vfr encoding at half full temporal resolution for the credits... just add the tdecimate line)
s=mpeg2source("whatever.d2v")
a=s.trim(0,31287)
#b=s.mcbob().selecteven().trim(31288,33096)
#b=s.tempgaussmc_beta2(edimode="nnedi").selecteven().trim(31288,33096)
c=s.trim(33097,34575)
a+b+c

For your other source, it can be #2 case (Double hard telecined encoded as progressive) since there should be no interlacing left after mode=1's field matching... you saw only 1-2 frames, but there could be more. I'd recommend trying mode=2 to see what happens with your clip.

[egrimisu]

Thanks, for the second source i tried mode 2 but everything is scruwed up; the interlancing frames that i told u about apears only at an interval of 30 seconds or more, its sure not double hard telecined. Thanks i'l try the scripts that you writen but i would liked more the mode=6 thing ;)


I cold post a sample later today or tomorow morning.

[thetoof]

Yes, I'd like to have a sample to try to understand why the ouput still has interlaced frames.

As I said, you don't need IVTC, so no mode of AnimeIVTC is appropriate for your source, unless at some point your stream was telecined.
If you absolutely want a function, try this
#DeintPart, helper function by thetoof
#
#Output = 1 (29.97fps) or 2 (59.94fps)
#Decimation = 1 (CFR, no decimation) or 2 (VFR decimation)
#Bob = 1 (mcbob) or 2 (tempgaussmc_beta2)
#Block, hybriddup, pass = like in AnimeIVTC()
#Denoise = true (denoise before calculating metrics for VFR decimation) or false (don't)
#pstartx/endx = frame # in the unprocessed source where the progressive parts start and end
#istart1/end1 = frame # in the unprocessed source where the interlaced section starts and ends

function DeintPart(clip s, int "output", int "decimation", int "bob", int "block", float "hybriddup", int "pass", bool "denoise", int "pstart1", int "pend1", int "istart1", int "iend1", int "pstart2", int "pend2")
{
block=default(block, 16)
hybriddup=default(hybriddup, 0.9)
pass=default(pass,1)
denoise=default(denoise,true)

a = (output==1) ? s.trim(pstart1,pend1) : (output==2) ? s.trim(pstart1,pend1).changefps("ntsc_double") : nop()
b = (output==1 && bob==1) ? s.mcbob().selecteven().trim(istart1,iend1) : (output==1 && bob==2) ? s.tempgaussmc_beta2(edimode="nnedi").selecteven().trim(istart1,iend1) : (output==2 && bob==1) ? s.mcbob().trim(istart1*2,iend1*2) : (output==2 && bob==2) ? s.tempgaussmc_beta2(edimode="nnedi").trim(istart1*2,iend1*2) : nop()
c = (output==1) ? s.trim(pstart2,pend2) : (output==2) ? s.trim(pstart2,pend2).changefps("ntsc_double") : nop()

spliced = a+b+c

return (decimation = 1) ? spliced : (decimation = 2 && pass == 1 ) ? spliced.TDecimate(mode=4, output="stats.txt",denoise=denoise, blockx=block, blocky=block) : (decimation = 2 && pass == 2 ) ? spliced.TDecimate(mode=5, hybrid=2, dupthresh=hybriddup, input="stats.txt", mkvout="timecodes.txt",denoise=denoise, blockx=block, blocky=block) : nop()}
Never tested because my comp is still not operational, but it should do the trick...

[egrimisu]

stil don't undestand why have you made mode 6 in animeivtc since in the manual is writen that mode 6 i used for hybrid + interlance of every frame, or i undestand wrong and my source is not 29.97fps progresive with 30i on every frame that suites to mode 6?
Thanks for the scripts' il try them tomorow when i'll arive home.

[thetoof]

hybrid = mix of film @ 24fps telecined @29.97fps and video @ 29.97

You = (at least for the sample you provided) 100% video @ 29.97fps i.e. no need to apply inverse telecine + the "original" framerate is 29.97, not 24, so there is no need to decimate anything

Mode=6 is NOT for 30p, it's for hybrid (mix of 24t and 30p) and you seem to have only 30p

[ettin]

Hi, I have a hybrid video of 24fps film telecined into 30i, that has fx of 30p over it, mixed in with native 30i footage, part of which had fx'd laid over it and exported out as 30p which then had 30i graphics laid over it; all of which was exported out to 30i uncompressed. I then tried encoding this out to 30i mpeg2 (using CCE SP), which looked decent on the computer, but soon as i burned and played it on a few different players the footage jumped, lagged, and had strange field ghosting effects. I tried many different encoder settings and found the problem to be that my footage was interlaced where deinterlacing got rid of the ghosting, lagging and jaggies but destroyed the quality and altered the colors slightly. I took special care throughout the project to keep my export settings as bottom field first, that being what my footage was telecined into, so i dont understand how it could be a field issue.

I studied your guides and reference materials and i cant make sense of half of what your parameters do or how they apply to my video, but there appears to be a consensus that animeIVTC is the best filter for my type of problem, so i'm here now looking for help. Im not looking for any specialized way of fixing this mess (ie. applying various filters to certain parts as animeIVTC offers), i just need a solution that will turn all of it into something decent. Im a bit of a noob at this script writing process, could you offer a simplified explanation at what i should do and how to do it? It doesnt matter what format or fps the final product is, i just want it all to look like it does before it gets burned to disc.

[ettin]

Thanks for the CBR suggestion, it got rid of the bad jumping, but theirs still field ghosting and nasty interlacing bars. Any idea when the next version of IVTC will be released?

edit: hey, what happened to your post?

[thetoof]

I deleted my post cuz I was confused by your request and I thought "oh, what the hell, I'll re-read tomorrow after a good night o' sleep and post something more coherent".

Anyways, here's what confused me:
You say you don't want adaptative filtering and that you want a single solution for a clip with multiple issue... and you already have a format that is DVD compatible so I don't know what could be causing your problems.
If you want to remove all the interlacing, you need to make your footage progressive, but for that, I have a few questions:
1 - By burning to disc, do you mean burning files (avi, mkv, etc....) on a DVD or burning a MPEG-2 NTSC DVD?
2 - Different players = DVD players or softwares?
3 - Can you identify the frame # where every section starts and ends (24t, 30p, 30i) to apply something appropriate for all of them?

By ghosting, do you mean blending? If yes, it's normal to see blends when you deinterlace without bobbing an interlaced source, unless you drop a field...

Also, I never suggested CBR (constant bitrate), but I said that for a DVD, you need CFR (constant framerate) @ 29.97 fps, so it doesn't make much sense that you don't care about the framerate and the format if you want to encode in something DVD compatible.
Many questions, lots of confusion... will come back after a few hours of good ol' shleep :p

[ettin]

oh CFR, well, you inadvertently helped anyways.

What i have is a 8bit uncompressed 29.97 interlaced (BFF) avi file that includes the aforementioned mess of native film and native DV, which both are interlaced, yet have had progressive effects (created and exported at 30p) placed over them, or been entirely converted to progressive via fx processing, and then had interlaced graphics placed over that. Its really too much to discern one part from another, so straight IVTC would cause problems with the native DV parts and probably the section with 30p effects that were laid over the 30i (telecined) film, and god knows what else.

So i have taken this avi, encoded it through CCE using the doom9 settings for interlaced footage and its yielded an unwatchable product. The ghosting i am referring to is an effect where the interlaced fields linger for a fraction of a second between frames giving whatever is moving a jerky tracer like effect (if you've ever seen a digital pan, its similar in the way it stair steps in a jerky manner). Its very slight but still very noticeable, and almost looks like frames are missing, but i cant tell if its missing frames, or just lagging and jumping. i would show you an example but its only showing up when i play it in on a tv, which to answer your other question, ive tested this on multiple hardware DVD players (the ones that connect to a tv) and this effect is persistent, despite the fact its not present in the mpeg2 file when i preview it in VLC on my computer. All the visual problems are consistent across the movie, so there doesnt seem to be any bias as to the native film or the native dv sections.

So i have no clue whats happening, but either VLC is fixing the problem before i can detect it, or all DVD players are incapable of playing my mpeg2 correctly. I was always under the assertion that you should never see interlacing when interlaced footage (on a dvd) is played on a tv, yet the interlacing (fields/frames lagging, combing, and jagged edges) is visible and quite horrendous.

Thats when i tried to deinterlace it and the problems went away, but i lost half the quality, so im inclined to think that interlacing is the problem but i need a better solution that wont destroy the quality. Whatever solution you might offer, it must be able to process the whole movie uniformly as i dont have any way to sort out what is what.

Thanks for helping, this is the last step and then i can finally be done.

[egrimisu]

i could't find tempgaussmc_beta2 only tempgaussmc_beta1. Know a place where i can find it?

[thetoof]

@ ettin
VLC is most probably deinterlacing on playback, so I guess you'd want to do it before. As you said, it's kinda weird to see interlacing on a TV... anyways, here's what I can suggest:
1 - You can either bob your movie and cut the temporal resolution by half (removes all the interlacing, but drops some info).
tempgaussmc_beta2().selecteven() # or selectodd()
mcbob().selecteven() # or selectodd()
2 - get AnimeIVTC, all the requirements and call
daa()
to remove all the interlacing and blend the frames instead of dropping half of them

@ egrimisu
Here (http://forum.doom9.org/showthread.php?p=1171286#post1171286). It's the modded version to work with the multithreaded mvtools (that you can get a few posts above)

[tetsuo55]

- Very good combing removal and anti-aliasing functions (can be called externally as nnediaa() daa() and maa())
No AA (http://img88.imageshack.us/img88/1294/aliaseddy2.jpg) // AA (http://img153.imageshack.us/img153/2540/antialiasingqb9.jpg)



this script looks nice, and so do the examples!

However have you noticed how in the AA screenshot a lot of detail is lost, especialy in the cockpit?

[thetoof]

Yeah, I know... it's the price to pay to remove such nasty aliasing. The other modes are not that bad regarding detail loss (this is the most aggressive).

[egrimisu]

maa is the best detail preserve? i made test but i could tell witch is the best detail preserve.

[RsVolcom]

Here, finally, are some samples I'm curious about:
http://www.megaupload.com/?d=IS0EAYFQ
http://www.megaupload.com/?d=6MZ3GJDN

What mode should I be using for these? In the documentation once Film/Video drops below 95% I should be expecting a 3:2 pattern, but there are only small sections of 3:2 in much of these sources. I want to pick 'case 2' due to 'no specific pattern' but I've only got 90% film.

[egrimisu]

Hi i runned the next scrip in analysis pass in virtual dub and after complete i got a 0 bytes stats.txt file. What did i wrong?

s=DGDecode_mpeg2source("D:\Work Hanbun no Tsuki ga Noboru Sora\Hanbun_no-Tsuki_ga_Noboru_Sora_06.DVD(MPEG2.WAV)[Misu]_Track1.d2v", cpu=0)
setmtmode(2)
a=s.trim(0,39115).changefps("ntsc_double")
b=s.tempgaussmc_beta1(edimode="nnedi").trim(39116,41370)
c=s.trim(41371,43219).changefps("ntsc_double")
a+b+c
#pass1
TDecimate(mode=4, output="stats.txt",denoise=true)
#pass2
#TDecimate(mode=5, hybrid=2, input="stats.txt", mkvout="timecodes.txt",denoise=true)
smartfade(dgm=true)
source = last
backward_vec1 = source.MVAnalyse(isb = true, delta = 2, pel = 2, overlap=4, sharp=1, idx = 1)
backward_vec2 = source.MVAnalyse(isb = true, delta = 2, pel = 2, overlap=4, sharp=1, idx = 1)
forward_vec1 = source.MVAnalyse(isb = false, delta = 1, pel = 2, overlap=4, sharp=1, idx = 1)
forward_vec2 = source.MVAnalyse(isb = false, delta = 2, pel = 2, overlap=4, sharp=1, idx = 1)
source.MVDegrain2(backward_vec1,forward_vec1,backward_vec2,forward_vec2,thSAD=400,idx=1)
fft3dfilter(sigma=0.8, bt=1, bw=32, bh=32, ow=16, oh=16, plane=4, dehalo=0, ncpu=2)
ttempsmooth()
gradfun2db(thr=1.2)
crop (4,0,-4,0)
spline36resize(720,480)


To use the multithead from mvtoos 1.9.6.2-1 do i need to do something or just copy the dll file in plugins and thats all.
If i change the version of TempGaussMC_alpha3.avsi to beta2 animeivtc shoq an error, sometging with mv_analysis multithread... or somethind :D , you sorely know better.

[Nightshiver]

That error happens a lot. You obviously can't use the beta2 because animeivtc hasn't been made compatible with it yet.

[egrimisu]

And how can i fix it (the stast.txt thing)?

[thetoof]

maa is the best detail preserve?
Don't choose AA for what it's supposed to do, use the one that looks better...
Here, finally, are some samples I'm curious about
You sent me d2vs... I need m2vs (select a section of with a lot of motion, save and demux video in DGIndex)
I want to pick 'case 2' due to 'no specific pattern' but I've only got 90% film.
I should have said "mostly film"... you most probably have a case 2 source (though I need the samples to confirm). The guide will be updated at the same time as the function (should be done in 2 days or so)
Hi i runned the next script in analysis pass in virtual dub and after complete i got a 0 bytes stats.txt file. What did i wrong?
Did you close vdub right after? I need to do that to get the metrics written in the stats.txt file, so maybe that's your issue, cuz I tested your script and it works fine.
Also, you should cut your script right after TDecimate and do a lossless rendering pass for the second pass. It's pointless to run filters after the analysis pass, since it's purpose is only to calculate which frames to decimate during the second pass.
To use the multithead from mvtoos 1.9.6.2-1 do i need to do something or just copy the dll file in plugins and thats all. If i change the version of TempGaussMC_alpha3.avsi to beta2 animeivtc shoq an error, sometging with mv_analysis multithread... or somethind :D , you sorely know better.
Replace your old mvtools with the new one and open beta2 with the notepad. Where you see function TempGaussMC_beta2 at the beginning, erase _beta2 and it'll work with animeivtc().

Edit : Oh, and egrimisu, what about that sample that got through mode=1 with interlaced frames?

[egrimisu]

The sample will be upped Today, I PROMISE!!!

[clone83]

can you please upload your files to a different site other then mediafire as i have been trying to download them and nothing is happening... i even tried signing up and that did nothing aswell. thanks in advance.

[RsVolcom]

Sorry about uploading the wrong samples. Here are three samples each from the first two episodes. There's not too many scenes with high motion, but these should work. Thanks again.

Ep01 - Sample01
http://www.megaupload.com/?d=BZN461ND

Ep01 - Sample02
http://www.megaupload.com/?d=YI9WYCOG

Ep01 - Sample03
http://www.megaupload.com/?d=8TNAVR6U

Ep02 - Sample01
http://www.megaupload.com/?d=RJFXDM6G

Ep02 - Sample02
http://www.megaupload.com/?d=C28107H5

Ep02 - Sample03
http://www.megaupload.com/?d=5KB6PV1S

[Nightshiver]

animeivtc(mode=2,aa=3,pel=4,search=3)

That worked well for me. You could also use tfm(order=1).tdecimate(hybrid=1)

[RsVolcom]

animeivtc(mode=2,aa=3,pel=4,search=3)

That worked well for me. You could also use tfm(order=1).tdecimate(hybrid=1)

Thanks for your effort I am trying this now. Could you please explain what 'pel' and 'search' do?

pel : Supersampling factor for motion vector calculation. Higner = better, but slower

* 1, 2 or 4 (default=2)

search : Search effort. Higner = better, but slower

* 2 (Logarithmic search, also named Diamond Search) (default)
* 3 (Exhaustive search)

[Nightshiver]

lol, apparently you answered you own question. You honestly don't have to increase the pel and search.

[thetoof]

can you please upload your files to a different site other then mediafire as i have been trying to download them and nothing is happening
Ok... never seen that before, but I'll put some mirrors in the next release.

@RsVolcom
dling your samples now
pel and search affect the quality of the motion vector search used for the framerate conversion after bobbing interlaced credits. It doesn't have any effect in mode=2 (in the next version, it will if you use a precision higher than 2)

[RsVolcom]

lol, apparently you answered you own question. You honestly don't have to increase the pel and search.hehe I was trying to show that I had read the documentation, while also asking for a dumbed-down version of what vector calculations were. I read a few other threads and I think I almost get it.

With mode=2 I get terrible jumping around of horizontal lines(on ep 2). I tried mode=2 with every aa=* option but they all looked bad. Here's the script you suggested, nightshiver:
http://www.megaupload.com/?d=CFQ4ORKB

The best results I had with episode 2 were using mode=1, aa=2:
http://www.megaupload.com/?d=87PJJODW

[Nightshiver]

That's actually odd because I didn't see a discernable 3:2 pattern.

[RsVolcom]

That's actually odd because I didn't see a discernable 3:2 pattern.No I definitely don't think there is one. The pattern seems very random.