Modern Anime IVTC

[Ranguvar]

I've been searching around, and the only information I've found on IVTC-ing anime is from at least four years back. One thing I keep finding, is that anime is apparently often hard to IVTC.

So I wanted to know: What high quality methods exist today for doing this? :)

[TheRyuu]

I've been searching around, and the only information I've found on IVTC-ing anime is from at least four years back. One thing I keep finding, is that anime is apparently often hard to IVTC.

So I wanted to know: What high quality methods exist today for doing this? :)

Yatta

[Chainmax]

YATTA is quite old too.

Thing is, Ranguvar, the theory on how to handle each case has not changed much, only the tools with which to handle them have somewhat evolved. McBob+NNEDI, TIVTC, Restore24 and MOmonster's restore pack are still the most widely used tools in their respective fields.

[MadRat]

It depends, if you just don't care about field blending and you want to get it done the fast, easy way and you have top field first (TFF) use...


deinted = tdeint(mode=2)
tfm(clip2=deinted)
TDecimate(mode=1)


If you're using YATTA with field matching you'll see the letters C, N and P. These stand for current field, next field and previous field. (There is also a U and a... I think it's a D isn't it? That's for matching the previous top field to the next bottom field or vice versa. Normally you don't use those very often but heaven only knows what you'll find on an anime DVD.)

For more information on the easy way out try

http://www.animemusicvideos.org/guides/avtechbeta/progressive.html

[TheRyuu]

YATTA is quite old too.

Thing is, Ranguvar, the theory on how to handle each case has not changed much, only the tools with which to handle them have somewhat evolved. McBob+NNEDI, TIVTC, Restore24 and MOmonster's restore pack are still the most widely used tools in their respective fields.

It may be old, but it's still (sort of) maintained and give me another program that ensures proper decimation on some anime dvd's that aren't soft telecine. :p

[Zarxrax]

Honestly, I've found that most recent anime IVTC pretty much perfectly.
It's the old stuff thats a pain :)

[mahsah]

I've found that the 2 pass VFR method with tivtc is great, especially if you arn't sure about framerates.

[martino]

I've found that the 2 pass VFR method with tivtc is great, especially if you arn't sure about framerates.
That "VFR" mode however makes one hell of a mess... In that case I think it's better to assume that it is all 23.976 and decimate, or it isn't and don't... :/

*points at most anime being 23.976 and 29.97*

[shirohamada]

dgindex --> forcefilm

XD

[dbzgundam]

Anime is essentially a format in its own if you ask me. Generationally speaking, we can separate the difficulty of anime into categories based on their age. Not just of the series itself, but of the transfer/DVD.

"Older" Shows: usually identify will anything that was cel-based, photographed onto film, and most likely hard telecined for the original TV masters and original DVD releases. A possible exception to this rule are extremely modern cel based shows (Big-O, Inuyasha, and anything remaining the 2000s), whose DVDs often have component transfers and very consistent telecines.

"New" Shows: Usually refers to anything in the 2000s up produced with computers. The problems of hard-telecined cel based shows have been trumped with the consistency of a digital workflow; HOWEVER, inconsistencies such as differences in programs, resolutions, framerates, special effects, and budget lead to, in my opinion, far worse telecine issues with CG anime.

To diverge for a moment, CG anime (not to be confused with CGI animation; I'm referring to 2D animation that is colored on a computer) CAN be (I'm not saying I know precisely how it's done) colored in one program and animated together in another. There's a ton of issues that come up in the editing stage though, where editors may even work in 60i (presumably after the animation itself has been telecined) and then integrate purely 60i special effects, credits, etc. There are even cases where a 60i pan is performed across the animation, presumably because the motion was already telecined.

This of course causes another huge problem: HD capability. Already there are mere upscales for Blu-ray anime discs. Yukizake's BD is an upscale (unsurprisingly from the "masters" of telecine, Gonzo) judging by the utter lack of detail and sharpness. The few pictures I've seen have shown combed frames in various places, which makes one wonder just how possible it is to easily resize these shows (presumably edited in SD... I'll get to why I can guess that later) to HD.

Take Full Metal Panic S1 (I refuse to buy the second and third season R1 DVDs due to horrible authoring). The DVDs are decent, with some artifacting, but the mastering is absolutely terrible. In any given episode, there is not only VFR, but 60i zooms, double telecines, and 60i fades. This would lead one to believe that at least the 60i stuff was added very late in production, most likely by the editor, and after the show was already output for SD. What will happen to this show beyond SD? The combs can be helped very manually, but the dissolves, pans, and zooms would have to be rerendered to look decent at all.

To get back on topic though...

"Old" Shows on Remastered DVDs: These are the best in my opinion. Since the older cel-based shows are entirely on film, most remasters (Evangelion being a nasty exception) are easy to IVTC, with some even being able to have Force FILM applied to them (Outlaw Star comes to mind). There *should* be no possibility for VFR in these shows.

"Really New" Shows: Since about 2005, MAYBE 2004, the mastering quality of CG anime has increased, admittedly. Gundam Seed Destiny for example, brought Gundam Seed's framerate to a solid 24fps (Gundam Seed was practically all 30p, with few exceptions), with one or two flashback sequences breaking the trend. Sunrise even rerendered several sequences (mainly the recycled OP animations) to playback in 24p. VFR is still abundant, but the break in framerates isn't as problematic as it once was, and 60i stuff isn't as common. Additionally, with many more shows being natively produced in HD, there is a push for high quality mastering.

There's my little rant, I hope I reduced someone's faith in anime DVDs.

[Zarxrax]

I also wanna throw in the fact that NNEDI seems to be awesome for deinterlacing anime. I've been using it as a postprocessor for TFM recently with great results.

[MadRat]

Do you have a code example for us, Zarxrax? :)

[Zarxrax]

Basically the same thing you posted above with tdeint, only replace the tdeint part with NEEDI()

[Terranigma]

(There is also a U and a... I think it's a D isn't it?

You mean U & B, and iirc, YATTA has yet to have support for them; just P, C, & N.

[DarkT]

Do you have a code example for us, Zarxrax? :)

Interp = nnedi(field=0)
deint = TDeint(mode=0,Order=-1,type=1,AP=55,edeint=interp,slow=2)
tfm(d2v=d2vpath,clip2=deint,order=1,Mode=5, pp=6, slow=2)

Something like that... It's a minimalisation from my Claymore script... Which I'm doing in a very lazy way, unlike the Eclipse guys ;).

Edit:

As per yatta... I never learend to use it :) - it's all down to override file with me :), same as yatta, only yatta supposedly can - in the hands of a master - make the work 3-4 times faster :).

Edit2:

Beware of NNEDI usage though, with some old stuff it can mess your stuff up, if the old stuff is funky...

Edit3:

Oh, my script is a bit different, got another extra step...

[Chainmax]

The results are even better if you include TMM in the mix:

IVTCing
For TFF clips:
AssumeTFF()
Interp = nnedi(field=1)
Deinted=TDeint(order=1,field=1,type=1,edeint=Interp,emask=TMM(order=1,field=1))
TFM(d2v=path_to_d2v,mode=6,order=1,PP=7,slow=2,mChroma=true,Clip2=Deinted)
TDecimate(mode=1)

For BFF clips:
Interp = nnedi(field=0)
Deinted = TDeint(order=0,field=0,type=1,edeint=Interp,emask=TMM(order=0,field=0))
TFM(d2v=path_to_d2v,mode=6,order=0,PP=7,slow=2,mChroma=true,Clip2=Deinted)
TDecimate(mode=1)



Deinterlacing
For TFF Clips:
Interp = nnedi(field=1)
TDeint(order=1,field=1,type=1,edeint=Interp,emask=TMM(order=1,field=1))

For BFF Clips:
Interp = nnedi(field=0)
TDeint(order=0,field=0,type=1,edeint=Interp,emask=TMM(order=0,field=0))



Bobbing
For TFF Clips:
Interp = nnedi(field=3)
TDeint(mode=1,order=1,type=1,edeint=Interp,emask=TMM(mode=1,order=1))

For BFF Clips:
Interp = nnedi(field=2)
TDeint(mode=1,order=0,type=1,edeint=Interp,emask=TMM(mode=1,order=0))

Of course a higher quality alternative to the deinterlacing and bobbing methods would be to use McBob&NNEDI+SelectEven (or was it SelectOdd? can't remember) and McBob&NNEDI respectively.


[edit] DarkT: NNEDI can cause problems? What kind and in which cases?
[edit2] forgot to put the d2v option in TFM's line, reading DarkT's script made me remember :)

[DarkT]

yarrrr, "PP=7" lol, not something I use much, mostly I give it pp=5... Or do I? Hmmm, don't rememebr, ahh, it's prolly the claymore thing... I opted for making tons of blends in it, also TDecimate wise :) (instead of goign VFR on it)...

As for what NNEDI can do... Well, from my BERSERK dvd, and a few others - it can "shred" the place of the motion into pieces :).

TFM would create a "ghost" like artifact, while using NNEDI creates mostly ok image, but can really "shred" it in places... Grrr, it prolly has to do with the fact that 1st we throw it back to Tdeint, which throws it to NNEDI... I checked and rechecked and tried to make it good and well, with one of my TRIGUN dvds, it was no go... BUT - the good news is that most people won't even notice it :) it happens rarely, and when it does it's just 1 frame here, 1 frame there ;).

Edit: Hmm, TMM? Never heared of it... *goes to google*
Edit2: Hey - could you give some more info on TMM? Can't find much of anything about it besides that it's used with Tritical's stuff lol ;).

[Chainmax]

We are both stumped then (;)), what is Claymore? :o

TMM is a filter that creates a motion mask for other filters to use. If I recall correctly, it was inspired by McBob's motion masking. Some more explanation about it is available here (http://forum.doom9.org/showthread.php?t=82264&page=49) and you can download it from here (http://bengal.missouri.edu/~kes25c/TMMv1.zip).

By the way, I use PP=7 because that's ELA interpolation, which according to the readme should smooth out jagged edges.

[Zarxrax]

I might be wrong on this, but I think the pp=7 is completely unnecessary since TFM is not doing the deinterlacing. I would think the tmm is unnecessary too since I believe tdeint can make its own mask.

[DarkT]

I might be wrong on this, but I think the pp=7 is completely unnecessary since TFM is not doing the deinterlacing. I would think the tmm is unnecessary too since I believe tdeint can make its own mask.

the clip2 parameter of TFM:

clip2 -

When using PP > 1 this parameter may be used to specify a clip to take deinterlaced
frames from instead of using one of TFM's inbuilt deinterlacing methods. If PP < 5
then frames will be taken as is from clip2, if PP >= 5 (i.e. using motion adaptation)
then TFM will build the mask as usual and only pixels in moving areas will be taken
from the frames in clip2. Don't specify anything for clip2 to use TFM's internal
deinterlacing.

Further more, the Edeint parameter of Tdeint:

edeint:

Allows the specification of an external clip from which to take interpolated pixels instead of having TDeint use one of its internal interpolation methods. If a clip is specified, then TDeint will process everything as usual except that instead of computing interpolated pixels itself it will take the needed pixels from the corresponding spatial positions in the same frame of the edeint clip.

Chain - as for claymore - it's only one of THE hottest animes of the past season :).

Edit:

"TFM will build the mask as usual and only pixels in moving areas will be taken from the frames in clip2."

I bet this somehow is related to NNEDI shredding stuff into pieces sometimes... My guts tell me so :).

Edit2:

"I use PP=7 because that's ELA interpolation, which according to the readme should smooth out jagged edges."

Lol, aren't you applying info from Tdeint readme on TFM? lol, does that work that way? :).

[Zarxrax]

Ah, hmm.
Well if TFM builds a motion mask, then what is tdeint doing?

[DarkT]

*glares at tritical/didee/other-people-who-speak-avisynth*

I'm not big on technical stuff :) I try 100 different things and decide by what looks better :) - I'm not even 100% certain WHAT a mask IS :). A MOTION-MASK? Please... :) :).

[Zarxrax]

Motion mask: Look at the areas of the frame that actually contain interlacing, and then you can deinterlace only these parts instead of the entire frame.

So, here's what I *think* I see going on in Chainmax's script:

TFM: matches fields, then for interlaced frames it builds a mask of just the interlaced areas, and then replaces those interlaced areas with the tdeint clip.

Tdeint: Just tells TMM to build it a motion mask, although it's apparently capable of building its own, although tfm already did it... and then instead of using it's own deinterlacing methods, it tells NNEDI to do the deinterlacing for it.

TMM: Just builds a motion mask...

NNEDI: Just deinterlaces.


It all sounds insane to me, but maybe there is a purpose to it :)