anime ivtc

[sugarkang]

hi everybody.
this is my first post, but i'm not a total gknot noob. i've ripped about 30 things with gknot so far, and am really impressed.

so far i've gone through all the gknot posts regarding anime, and i've also read the *deinterlacing sticky/FAQ* several times and i'm still unclear about some IVTC stuff.

i'm using gordianknot 26.1 with divx pro 5.02
i've also downloaded and replaced decomb.dll with version 3.91
i'm all up to date right?

i'd really appreciate some help because i'm about to start ripping the entire robotech series [NTSC] which is 85 episodes, and want to ensure pretty good results. it doesn't have to be absolutely perfect. just no interlacing, and no ringing :)

anyway, here's what i'm stuck on:

ISSUE #1 - the deinterlace FAQ says to study the .avs carefully, and to know all the possible settings for IVTC. also, use a deinterlacer in conjunction with IVTC. i thought only one or the other was possible. so do i just edit the .avs manually to run ivtc and deinterlace? i know this was touched upon in a couple threads, but i can't really tell what is good information and misinformation.

ISSUE #2 - the deinterlace FAQ didn't want to be partial to any IVTC except IVTC2.2 is supposed to be good for anime. isn't there an ivtc 4.0 now also? i'm confused what to use, and whether to use this in conjunction with a deinterlacer.

please help. i know how to run gknot, i've read the FAQ many times, and i've also searched through the forums for all the posts, but i'm still confused.

P.S. - manono... you work too hard man! you helped me with a lot of issues. thanks.

[manono]

Hi and welcome to the forums-

you helped me with a lot of issues. thanks. You're welcome, but if you're referring to the tutorial, it was a true joint effort. But some of what you're asking refers to things I wrote, so if it was unclear, then it's my fault, although I purposely evaded some IVTC issues to avoid stirring up a hornets nest. DJ Bobo has scolded me already for not promoting IVTC2.2 enough. :)

First, don't confuse IVTC with deinterlacing. They are 2 different processes. IVTC (inverse telecine) is the process of undoing the telecining. It reconstructs the frames and tosses out the duplicates to return the material to it's original 23.976fps. Deinterlacing blends or interpolates any remaining interlacing that slips through the IVTC process. That was mentioned at least twice in the tutorial before my section. Have a look at the little picture to see what happens during the telecine process. The bottom picture is after telecine, and the top picture is the original frames, as well how it should again look after IVTC.

So, Decomb's Telecide().Decimate(5) combo performs IVTC and deinterlaces, as FieldDeinterlace() is built into it by default. IVTC2.2 only performs IVTC and if you want to deinterlace as well (recommended), you'll have to add Decomb's FieldDeinterlace() or some other deinterlacer. If you decide to go with Decomb for your project, then read and reread the Decomb Help HTML to learn all the various combinations and permutations. But, with any luck, what works for one episode should work for the other 84.

i'm all up to date right?

Not really-Go into the AviSynth Forum and get the latest version. It's up to Decomb4.00RC6, I think. Just replace the Decomb.dll in GKnot with the most recent one, and put the help file in with the other Docs. Here (http://forum.doom9.org/attachment.php?s=&postid=188252) it is. I saved you the work.

I don't know anything about an IVTC4.0 (maybe you're thinking of Decomb4.0). IVTC2.2 is a different IVTC. You'll have to get the .dll from the downloads section, dump it into GKnot, and then Load the Plugin and add IVTC(44,11,95) into your script. Nevermind, Here's IVTC2.2 (http://www.doom9.org/Soft21/Filters/IVTC22.). I saved you some more trouble. Flattery will get you everywhere.:) Good Luck with Robotech.

[sugarkang]

thanks for the help as always manono.

i tried IVTC2.2 on a ninja scroll sample 3 mins long with heavy movement. i tried the following:

1. IVTC2.2: 32,11,110 [according to the ultimate ivtc thread]
2. IVTC2.2: 44,11,95
3. decomb 4.0 - [gknot default]
4. decomb 4.0 - post=true,blend=false,guide=1,gthresh=44,threshold=20,chroma=true,dthreshold=12
5. decomb 4.0 - [gknot default, but postprocessing disabled]

conclusion:
1. both ivtc2.2 left a lot of combed material, but i guess that could be deinterlaced.
2. decomb without postprocessing leaves less combed material than ivtc2.2
3. decomb 4.0 default was the best. [for this test]

anyway, i'd like to know which one you use more... ivtc or decomb? i know that one will work better than the other in some cases and vice versa... but i'd still like to know which one you personally use more often for MOST jobs, and with what settings... if you're customizing.

btw... "IVTC.dll" is the upgrade to "InverseTelecine.dll" so i can just rename IVTC.dll --> InverseTelecine.dll so that i don't have to manually type in stuff each time i edit the script?

[hakko504]

Originally posted by manono
I don't know anything about an IVTC4.0 (maybe you're thinking of Decomb4.0).Actually he's right, there is an IVTC4 moldule available. The consensus in the AVIsynth forum was that it can't beat Decomb at the moment (D3.2).


I should like to point out that while we use Decomb in the guide the other IVTC modules can do just as well in most cases and even better than Decomb in some. We chose to use Decomb becase it is a module we assumed everyone had (through GKnot) and becase it can do so many things besides IVTC, like pure de-interlacing and frame decimation (for 18fps). It is by far the most versatile IVTC module around. It also has very descriptive parameters, thus it is quite easy to read from a script example what happens in the script. This way we could write the guide and only make references to one module. On the other hand the guide was written when Decomb v3.91 was available and since neuron2 is striving for perfection:), v4.00RC6 might be very hard to beat in ANY case.
Still, if you want the best possible rip you must be prepared to try a number of different combinations of IVTC methods and parameter values to those methods.

[jggimi]

Good question, sugarkang, about which IVTC when.

In my case, if the content is telecined, but can't be Force FILMed, I'll always start with:

Telecide(post=false)
Decimate(cycle=5)

I'll examine a scene frame-by-frame in Vdub, to see if Telecide without de-interlacing will work to reassemble the frames. If so, this will look as clean as a Force FILM, and will save some processing time.

If after reassembling frames, de-interlacing is required, I'll remove the post=false ... the Gknot default when you select IVTC.

If I don't particularly like what I see, it may be because I've got a non-standard telecine. If so, I might have to adjust the Decimate options, and treat it like a hybrid.

After playing with DeComb, if I'm not satisfied -- it's only happened once, so far -- I'll give GreedyHMA a try, adjusting options and previewing results in Vdub.

I've found it is a little easier to see the preview by having no resizing filter in the test .avs script. Either I build the .avs manually, or I comment out resizing in a Gknot-built script.

[manono]

Hi-

there is an IVTC4 module available.

Of course you're right hakko504. I had forgotten about Daxab's IVTC4. sugarkang's been digging through some old posts. Sorry, Daxab-didn't mean to insult you.

sugarkang-you didn't IVTC a Ninja Scroll .avi by chance, did you? You have to use .vobs. If you IVTC an .avi, the field information is already gone, and it doesn't work.

As for what I use, about 95% of the time it's Decomb. I work with all kinds of movies, and for such things as repairing crummy PAL to NTSC conversions, hybrids, and silents when you're trying to get the original fps back out of them, it's the only game in town. I've used IVTC2.2 a few times with anime, and used to use GreedyHMA and TMPEGEnc a lot. But as hakko504 says, others an be equally good in many cases.

"IVTC.dll" is the upgrade to "InverseTelecine.dll" so i can just rename IVTC.dll --> InverseTelecine.dll so that i don't have to manually type in stuff each time i edit the script?.

Good One-I hadn't thought of that. I've been typing it in when I use it. Since you already have the InverseTelecine line in the GKnot generated .avs, then it'll save some time, and I don't see why you can't do it that way.

Edit: And I agree that Decomb4.0 is a major improvement. The defaults are very good now, and no longer do I have to tweak the Gthresh settings.

[sugarkang]

sugarkang-you didn't IVTC a Ninja Scroll .avi by chance, did you? You have to use .vobs. If you IVTC an .avi,

i guess it's better to say for future noob readers that you don't IVTC anything unless it's 29.97NTSC. so no, i only IVTC'd the original vobs.

the thing that always confused me most about these framerates is that dvd2avi sometimes was a little buggy and reported 23.97 when i hit F5. clearly that's wrong. you would never IVTC 23.97.

i know now that all normal region 1 [NTSC] is 29.97 and you either have to:

1. IVTC then deinterlace if there are any combed frames left.
OR
2. just leave it at 29.97 if it's a purely interlaced source, and just deinterlace without IVTC.

I figure i spent about 10 hours to learn those two points.

[hakko504]

Originally posted by sugarkang
the thing that always confused me most about these framerates is that dvd2avi sometimes was a little buggy and reported 23.97 when i hit F5. clearly that's wrong. you would never IVTC 23.97.DVD2AVI will report 23.976 if Force Film is selected.

[sugarkang]

DVD2AVI will report 23.976 if Force Film is selected.

mystery solved :)

so i just redid ninja scroll due to jerky panning, and used the following:

Telecide(Post=true,Blend=false,Guide=1,Chroma=true)
Decimate(mode=2)

this cleared up the bad panning, but it took 12hours to encode on my athlon 1700+, as opposed to the normal 4 hours. i was hoping you could tell me whether this is due to the telecide options or decimate mode=2.

also, in the decomb4.00 help file, it says for decimate mode=2...

Note that for the most reliable IVTC of 3:2 pulldown material, you should use pattern guidance together with this mode of decimation.

what the heck is 3:2 pulldown and how do i know if i have that?

i did a forum search for "pulldown" and someone posted a link for:
http://www.doom9.org/index.html?/synch.htm

this only has information on 2:3 pulldown which is telecine. so is 3:2 pulldown just another way of saying IVTC? i don't think so. what's going on?

EDIT: also, if i edit the .avs manually, i can't get IVTC--> correct frame count right? so the movies won't be the size i want it to be? how do i get around this?

[Hiro2k]

I always understodd 3:2 pulldown was the same as saying IVTC since they both mean to bring the 29.97fps back down to 23.97. Look into this link about the IVTC correct frame count.

http://forum.doom9.org/showthread.php?s=&threadid=34752

[manono]

Hi-

OK, you didn't IVTC the Ninja Scroll.avi. The reason I had asked was that your results were strange. Nothing hard about IVTCing Ninja Scroll, and any IVTC should handle it easily. So you had made a different beginner's mistake by IVTCing on top of Force Film? Live and Learn. I'm not sure what you're comparing the encoding times to-the IVTC after Force Film with no deinterlace? That one would have been at 19.980fps and of course much faster than 23.976fps. When comparing to Force Film, IVTC slows you down some. Adding deinterlacing adds 30% or so to that (although interpolate is supposed to be slightly faster than blend). I guess the rest of the slowdown is due to Mode=2, but I haven't done any speed tests on it. It does sound like a big increase in time, though.

what the heck is 3:2 pulldown and how do i know if i have that?

It refers to the telecine process of increasing the frame rate from 23.976 to 29.97fps. 3:2 pulldown is used to mean the same thing as 2:3 pulldown, although in almost all cases 2:3 is the proper term. Rarely do they start on the second frame. In short, if your material started life as Film (as opposed to video or pure interlaced material), then a 2:3 pulldown has been performed. From our Tutorial (http://www.doom9.org/ivtc-tut.htm),:

If a frame consists of two fields, top (t) and bottom (b) and the original sequence is 1t1b 2t2b 3t3b 4t4b, then the telecined sequence would be 1t1b 2t2b 2t3b 3t4b 4t4b (commonly called 2:3 pull down because of the alternating 2 field, 3 field progression).

So, you have 2 frame #1 fields followed by 3 frame #2 fields, followed by 2 frame #3 fields followed by 3 frame #4 fields, hence 2:3 pulldown. The fields as displayed on an NTSC television follow a 2,3,2,3 pattern. Here's a Link (http://www.commotionpro.com/support/white_papers/3-2.html) that explains some more (it's also where I got the picture, and note that he's really illustrating 2:3 pulldown but, as I said, the 2 terms are used interchangeably). Here's another link (http://www.projectorpeople.com/news_info/pulldown.asp). Here's One (http://www.zerocut.com/tech/pulldown.html) that explains why 2:3 pulldown is usually the correct term (and is where robshot got his picture). And undoing the 2:3 pulldown is sometimes referred to as 2:3 pullup (we call it IVTC)! See this Link (http://www.highend3d.com/rosetta/?cat=1) for confirmation.