Cartoon cleanup

[HyperSpad]

Hey everyone, I'm still sort of new at this, but I've had my experiences with a slew of filters and have read through quite a few threads on this board, so i thought it was time to submit a post tailored to my situation and hopefully get some feedback on actions I can take towards a solution.

Tools being used: VirtualDub and Avisynth (mostly avisynth though)

First, let it be known that I'm primarily working with cartoons, not anime, cartoons. Mainly the ones found on cartoon network and/or adult swim, just to give an idea. (Examples include: Teen Titans, Justice League Unlimited, Harvey Birdman, Aqua Teen Hunger Force, and Family Guy)

2 big areas i'm currently struggling to get down...
dotcrawl/cross luminance and deinterlacing

Dot Crawl:
When I try to combat dot crawl, it seems like a battle of not getting rid of all the dot crawl VERSUS ending up with a perfectly "solid" still picture, but with ghosting in motion areas. I've looked through quite a few posts here on the subject, and then I made the smart decision of actually registering for the forum (recently).
I originally tried the "Dot Crawl Comb Filter" for VDub, but that wasn't powerful or customizable enough for my needs. Then I discovered the power of avisynth. I have fully experimented with guavacomb and no matter how much i tweaked it, I couldn't fully fix the problem (always seemed to encounter the problems I listed above). Then I found DeDot and ran it at the default settings and I think I like this one the best so far. It seems to do the best job, but I'm still getting some traces of noise. It mainly appears:

1. In areas where two lines (or curves) join at a very small angle (the distance between them goes from a few inches to a few pixels).
AND
2. Between two solid lines/curves that are only a few pixels apart but have a solid color in between them (I can post screen shots or even short clips if you need to see what I'm talking about). Is there a better filter I can use to eliminate dot crawl? Or how about another filter I can run after Dedot to clean up some of these extra imperfections? A good blurring filter, possibly?

And for the record, if I can, I want to try and avoid using sangnom (it's powerful, but it seems like it introduces too many other problems).


Deinterlacing:
I'm only getting small parts of my video interlaced, so it's "isolated interlacing", I guess. Little things need to be deinterlaced, like the movement of a hand or the turning of a character's head.

I'm, of course, open to suggestions regarding which plugin to use, but for now I feel "LeakKernelDeint" is doing a pretty good job.

The biggest problem I have with it is that it makes the black outlines around the "characters" too unsharp/pixelated when they move (the edges of solid color areas are a bit pixelated too). It's like using mspaint instead of photoshop to draw lines and curves. I'm happy that it doesn't leave any other artifacts alongside the interlaced areas (like I've found with other filters), but the lines definetly look like they can use some anti-aliasing or smoothing or something. Any good recommendations for anti-aliasing, smoothing, or blurring filters that can help me out here? (once again, I can post some screen shots or short clips if needed for clarification).

I wouldn't mind if there was a bit more blurring in areas of motion, but a deinterlacing filter that uses blending is just going too far!


Thanks in advance,
-HYPERSPAD

[Guest]

For dot crawl, capture using S-Video instead of composite video. Eliminate the problem at its source!

[MOmonster]

For the deinterlacing, just try tdeint. I think tdeint(type=3) works good for Cartoon, the ela algo makes the edges less steping.
I'm only getting small parts of my video interlaced, so it's "isolated interlacing", I guess. Little things need to be deinterlaced, like the movement of a hand or the turning of a character's head.
This sounds more like telecined and not real interlaced material, maybe you should also try tfm(post=7) (I think).

[mg262]

Cartoons are pretty much all I deal with too, although I mostly like stuff from the 80s -- nice to find someone with the same interests. Most of the stuff that people suggest for anime works reasonably on toons. Like @MOmonster says, you may just need to inverse telecine... I've yet to find a cartoon that is actually interlaced. But it may be telecined or it may be that fields are shifted so you get the top field from frame n together with the bottom field from frame n-1, making it look interlaced. Can you post a short segment?

[Chainmax]

First of all: are you capturing from TV or video? If you're capturing from video, then do as neuron2 says. The better the source the better the result. If you are stuck with TV capture through RF, then I'd suggest you to use DeDot() (http://nullinfo.s21.xrea.com/#DeDot_YV12) in this way:

AVISource("X:\wherever\blahblah.avi")
ConvertToYV12(interlaced=true)
DeDot()
your filters here

I too concur that you're likely getting telecined content rather than interlaced. Read the excellent documentation that comes with Decomb (http://neuron2.net/decomb/decombnew.html) in order to learn how to tell interlace from telecine. For IVTC, you can try either Decomb (http://neuron2.net/decomb/decombnew.html) or TIVTC (http://forum.doom9.org/showpost.php?p=678155&postcount=254). For Deinterlacing, try TDeint (http://forum.doom9.org/showpost.php?p=653978&postcount=213) like MOmonster said (why use ELA2 and not plain ELA? I'd assume slower is better...).

[MOmonster]

why use ELA2 and not plain ELA? I'd assume slower is better...
Of course type=1 will be also a solution (maybe a little bit better quality), but my eyes don´t see any quality differences. I just like to choose the faster way. :D

[HyperSpad]

First of all, just to resolve some quick concerns:

neuron2: As nice as that would be, I'm kinda stuck with capturing analog tv. I just love the thrill of finding the right filters to combine to end up with something better than I started with. Don't you? :D

mg262: What's up? You ever hang around in 80s-cartoons on the DejaToons network? I posted some examples.

Important discovery: ConvertToYV12(interlaced=true) fixed the anti-aliasing problem I was having on static areas (which I stupidly originally blamed on interlacing). Some of the edges of solid color areas I was ranting about before are fixed now. Thanks a lot Chainmax


http://www.ultronator.com/doom9/slade_bad.jpg
Without using ConvertToYV12(interlaced=true)

http://www.ultronator.com/doom9/slade_good.jpg
Using ConvertToYV12(interlaced=true)


Now I was gonna post clips, but they were like 200mb each and I didn't want to compress them. I felt that if I made them too short, it would've been too hard to see what I was talking about anyway. That, and I ran out of time. Maybe I'll post some tomorrow.


I only got to experiment with TDeint. I think these screen shots actually do a pretty good job of illustrating how each different mode worked (pay attention to the outlines)...

http://www.ultronator.com/doom9/raw.jpg
Raw

http://www.ultronator.com/doom9/ela.jpg
ELA: TDeint(order=1, type=1, sharp=true)

http://www.ultronator.com/doom9/kernel.jpg
Kernel: TDeint(order=1, type=2, sharp=true)

http://www.ultronator.com/doom9/ela-2.jpg
ELA-2: TDeint(order=1, type=3, sharp=true)

None of the results were completely perfect, but at least the lines aren't as jagged as they were before. I'm thinking maybe there's a way to fill in the gaps left in the black lines now (most noticeable on ELA-2), possibly with flaXen's Cartoon Tool (vdub filter) or some other filter I'm not yet familiar with yet. Any suggestions on this? Which setting do you guys think was the best? (if any)


I'm still looking forward to trying the other filters you guys posted. Tomorrow I'll play around with Decomb and/or TIVTC.

Chainmax: Why would I want to use a combination of Decomb and TDeint?


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

One of the other things that's bugging me right now, is this extra whiteish/clear noise that I'm not quite sure can be classified as dot crawl, cause it doesn't seem to go away with any of the dotcrawl filters (maxed out). It occurs between close together black lines (cartoon outlines). I'll have to post a clip of that tomorrow sometime though, cause it can't be seen in a screen shot.


Hopefully I'll have more time tomorrow.

[Chainmax]

...

Important discovery: ConvertToYV12(interlaced=true) fixed the anti-aliasing problem I was having on static areas (which I stupidly originally blamed on interlacing). Some of the edges of solid color areas I was ranting about before are fixed now. Thanks a lot Chainmax

...

None of the results were completely perfect, but at least the lines aren't as jagged as they were before. I'm thinking maybe there's a way to fill in the gaps left in the black lines now (most noticeable on ELA-2), possibly with flaXen's Cartoon Tool (vdub filter) or some other filter I'm not yet familiar with yet. Any suggestions on this? Which setting do you guys think was the best? (if any)


I'm still looking forward to trying the other filters you guys posted. Tomorrow I'll play around with Decomb and/or TIVTC.

Chainmax: Why would I want to use a combination of Decomb and TDeint?


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

One of the other things that's bugging me right now, is this extra whiteish/clear noise that I'm not quite sure can be classified as dot crawl, cause it doesn't seem to go away with any of the dotcrawl filters (maxed out). It occurs between close together black lines (cartoon outlines). I'll have to post a clip of that tomorrow sometime though, cause it can't be seen in a screen shot.


Hopefully I'll have more time tomorrow.

Actually, the ConvertToYV12 line was meant to be used along with the DeDot (dotcrawl remover) one, but it's good to see it helped with that too :). By the way, I didn't say "use a combination of Decomb and TDeint", I meant

For IVTC, you can try either Decomb or TIVTC. For Deinterlacing, try TDeint

You should really check if the source can be IVTCed, that deinterlacing test might not have been needed.

[mg262]

Cartoons really really are unlikely to be interlaced. If we can see 10 consecutive fields or so, even at a relatively low resolution, then were more likely to be able to say something concrete. I might edit in a link to somewhere where I did this in a few minutes, but using this on your source should do it:

reduceby2
#split the frame into fields and stackthem on top of each other
separatefields
stackvertical(selecteven, selectodd)
#then make sure we can see several frames at once
a = selectevery(5, 0)
b = selectevery(5, 1)
c = selectevery(5, 2)
d = selectevery(5, 3)
e = selectevery(5, 4)
stackhorizontal(a,b,c,d,e)

Open that up in VirtualDub and scroll until you find a section with lots of foreground animation (so different frames look different), resize to a sensible size and post that picture. (You might have to put in a crop beforehand if the final resizing makes the differences between frames too small).
mg262: What's up? You ever hang around in 80s-cartoons on the DejaToons network? I posted some examples.My main interest is in taking old cartoons like Vicky the Viking, Belle and Sebastian, Jayce and the Wheeled Warriors and dubbing them with foreign DVDs. I've looked at/posted in a lot of different places over time, but I don't really hang around anywhere but here, because this is the place with the know-how (at least AFAIK -- I don't mean to be rude about anywhere else). But most people interested in animation here are into anime, which is (technically) different from the kind of stuff I deal with in some subtle but important ways. So it's always nice to see someone into cartoons!

edit: @neuron2 was talking about analogue TV... just with a different connector to the one you're using. You would probably need a S-VHS VCR to use that kind of connection though. But the differences in quality are massive.

here is that link...its not quite like what you should get, but it's similar. In your case we would need the animation to be changing faster though.

http://forum.doom9.org/showthread.php?p=604415#post604415

incidentally, you don't have to post screenshots; VirtualDubMod as an option on the edit menu to copy the current frame into the clipboard, and VirtualDub probably has as well.

[HyperSpad]

First off, I'm not sure if this has anything to do with what was said so far, but the source is NTSC analog cable, just to set the record straight. NOT PAL.



Open that up in VirtualDub and scroll until you find a section with lots of foreground animation (so different frames look different), resize to a sensible size and post that picture. (You might have to put in a crop beforehand if the final resizing makes the differences between frames too small).


Here's two screen shots taken (using the helpful "Copy source frame to clipboard", was the Video menu, btw) in VirtualDub when there's a lot of motion in the foreground. I think this is what you were looking for. If not, let me know.

http://www.ultronator.com/doom9/sample1.jpg
As her fist pound's the table, watch the moth larva fly up into the air

http://www.ultronator.com/doom9/sample2.jpg
The hand reaching for the pizza

Hopefully these can help lead to some conclusions about the state of my source.

I'm gonna go do some experimenting with the Decomb pack on my own now to see if I can further unravel the mystery behind the telecining. I will most likely post again within the next few hours.


Oh yeah, and one more thing:

Actually, the ConvertToYV12 line was meant to be used along with the DeDot (dotcrawl remover) one...

I actually did use it with DeDot(), like you suggested, I just forgot to mention that in my last post. Dot crawl is still gonna need a bit more work, but for now I'm concentrating on deinterlacing/telecining.

[DarkNite]

As nice as that would be, I'm kinda stuck with capturing analog tv.

Are you saying you don't have a way to use S-Video out from your source? I know my home office's cable box has an S-Video out and it's just an old ("Yeah, but this is the newest of the old boxes" - Cable Guy) analog box.

[mg262]

Duh duh. duh. I'm a total idiot. I added the reduceby2 after writing the script + pasted it in the wrong place...

#split the frame into fields and stackthem on top of each other
separatefields
stackvertical(selecteven, selectodd)
#then make sure we can see several frames at once
reduceby2
a = selectevery(5, 0)
b = selectevery(5, 1)
c = selectevery(5, 2)
d = selectevery(5, 3)
e = selectevery(5, 4)
stackhorizontal(a,b,c,d,e)

I'm really sorry about that. :stupid: I'm with totally moronic. Adding reduceby2 (half horizontal and vertical resolution) at the start simply blends together the two fields. It is needed in the script because VirtualDub won't cope with the picture otherwise (too wide) but it has to be put in after the fields are separated.

(The results that are coming out do look compatible with telecine to me though.)

[HyperSpad]

Are you saying you don't have a way to use S-Video out from your source? I know my home office's cable box has an S-Video out and it's just an old ("Yeah, but this is the newest of the old boxes" - Cable Guy) analog box.
I don't have a cable box. I guess if i went out and bought a vcr that has s-video, (not sure if I have one), and hooked the cable up to it and ran the s-video cable from the vcr to my computer's tv card. But I don't really wanna spend the money on such things. Maybe i'll try that eventually, but for now, I wanna see how much I can take care of with filters. I'm gonna be trying to get rid of dotcrawl next, seeing as I think I'm just about done with deinterlacing...


Duh duh. duh. I'm a total idiot. I added the reduceby2 after writing the script + pasted it in the wrong place...


Oh well, mistakes will happen. I tried the new settings and replaced the screen shots in my previous post (above) with the new results.

I read through Decomb's manual and tutorial and I'm almost positive I was getting telecining and no real interlacing! I also learned that my clip can be catagorized as having mostly "3:2 pulldown" and I think the telecining problem is completely gone now! Here's what I used...


AviSource("V:\record\the_end_new_clip.avi")
ConvertToYV12(interlaced=true)
# top field first, order=1
# we have mostly 3:2 pulldown, guide=1
Telecide(order=1,guide=1,post=0)


Note, that I didn't Decimate() it. I don't think it's really necessary. If there's duplicate frames, they aren't causing me any trouble, (I've watched my results several times), and I don't wanna mess with the frame rate. The tutorial warned that if my source wasn't fully 3:2 progressive, then I might get jerky motion if I decimated, so I'm not gonna bother.

[Chainmax]

Are you saying you don't have a way to use S-Video out from your source? I know my home office's cable box has an S-Video out and it's just an old ("Yeah, but this is the newest of the old boxes" - Cable Guy) analog box.
Well, at least in my case, the cable goes straight to the TV. There are no intermediary boxes except for one you can buy in order to watch extra channels (like the pr0n, oldies or exclusive sports ones).

[Chainmax]

I read through Decomb's manual and tutorial and I'm almost positive I was getting telecining and no real interlacing! I also learned that my clip can be catagorized as having mostly "3:2 pulldown" and I think the telecining problem is completely gone now! Here's what I used...


AviSource("V:\record\the_end_new_clip.avi")
ConvertToYV12(interlaced=true)
# top field first, order=1
# we have mostly 3:2 pulldown, guide=1
Telecide(order=1,guide=1,post=0)


Note, that I didn't Decimate() it. I don't think it's really necessary. If there's duplicate frames, they aren't causing me any trouble, (I've watched my results several times), and I don't wanna mess with the frame rate. The tutorial warned that if my source wasn't fully 3:2 progressive, then I might get jerky motion if I decimated, so I'm not gonna bother.

If it is telecined, then not decimating is a VERY BAD IDEA. Why? Simple. Telecined material is originated from a FILM source that was shot/drawn/whatever at 23.976 frames per second. Each frames is divided into fields (one having the odd lines, one having the even ones). Some of these fields are duplicated and recombined so that a framerate of 29.97 fps is achieved. So, a telecined stream does not have more frames than the film source. What field matching (telecide, in your case) does is simply undo the recombining, which yields the original frames plus the duplicates. Therefore, by not eliminating those duplicates (using Decimate) you will only waste bitrate on duplicate frames and have a stuttery video. This brief explanation is not 100% accurate but I hope it's clear enough. For more info on this, check out this article (http://www.doom9.org/ivtc-tut.htm).

Also, you didn't interprete Decomb's instructions correctly. When a source is "mostly 3:2 pulldown" it doesn't mean "mostly progressive", it means that there were some irregularities in the telecining process, but the source is still completely progressive and the original FILM rate can be achieved with IVTC (more info on that can be found on the article I linked to).
One more thing, setting post=0 is not a good idea IMO. After IVTC, some residual combing might slip by, and postprocessing is meant to eliminate it, so you should not disable it.

[HyperSpad]

So Interlacing/Telecining was one of the two original problems that caused me to start this thread. Now that that's under control, I wanna ask some questions about dot crawl. I know so far the general opinion is that I should get s-video input, but I think that with the right combination of filters, the dot crawl can be combatted!

Now before you go yelling at me claiming I haven't checked any of the other posts on these forums, let me say that I've checked quite a few. And the result always comes out to be the same..."Use guavacomb, that'll fix everything!"...or..."use DeDot, that filter is soo awesome!" Well, so far, none of those two are flawless, so I continue with my post...

I wanna split up the problem into a few different areas, one of which I don't think is truly dot crawl, but you can be the judge of that...

1. Border Dot Crawl: Dot crawl that accumulates on the left and right edges of the frame of video:
http://www.ultronator.com/doom9/edge_dotcrawl.jpg

2. Strange white dot crawl (or remnants of dot crawl):
http://www.ultronator.com/doom9/white_crawl.jpg
(Follow the red arrows)

3. Any other Dot Crawl found:
http://www.ultronator.com/doom9/dotcrawl.jpg
(notice the top rim of the table and the Cartoon Network logo)

And, I just couldn't resist posting this one. How's that for a headache?
http://www.ultronator.com/doom9/couldnt_resist.jpg

Here's the script that produced these screens:

LoadPlugin("C:\Program Files\AviSynth 2.5\plugins\DeDot.dll")
avisource("V:\record\the_end_new_clip.avi")
ConvertToYV12(interlaced=true)
Telecide(order=1,guide=1,post=0)
DeDot(20,20,15,5)


I guess DeDot() can't take care of everything after all. It got rid of some dot crawl, and minimized other parts, but it's not enough by itself to get a nice clean picture. I've tried playing with the settings for DeDot() a bit, but they seem to cause ghosting if I set them too high. I've played around with guavacomb significantly more than dedot, but I don't see either of them being able to fix the problem without causing significant amounts of ghosting.

So please, feel free to share your strategies on one or more of the above "situations."


Thanks again everyone
HYPERSPAD


Btw, I'm in the same situation as Chainmax when it comes to how my cable is setup.

[HyperSpad]

If it is telecined, then not decimating is a VERY BAD IDEA...


Ok, I'll try decimating. It couldn't hurt to try. I'll also see what enabling post processing will do.

[Chainmax]

Hyperspad, DeDot is meant to be used before IVTC or deinterlacing, that's probably the reason why it's not being so effective. Like I already said, the correct way of using DeDot is the following:

AVISource("X:\wherever\blahblah.avi")
ConvertToYV12(interlaced=true)
DeDot()
your filters here

...which uses the default settings. Don't bother toying with the settings, everyone here uses the defaults (which IIRC are what you used anyway) since the documentation is in japanese and no one knows exactly how to tweak them.

Now, I have no idea what the border artifacts in the first two pictures are, but I'd say they aren't dotcrawl. You can just crop then away. Then again, I'm completely ignorant on the issue so I'll let someone else talk about that.

The artifacting pointed on the red arrows and especially in the last pic do seem like dotcrawl to me, so try using DeDot correctly and report back. If that doesn't solve the issue then I'm out of ideas (again, I haven't dealt much with dotcrawl so I can't identify with certainty).

[mg262]

As you figured out, those pictures clearly indicate pulldown. For future reference, the mix of progressive looking frames (top field and bottom field match) and interlaced looking frames (top field and bottom field mismatched) is characteristic of pulldown. Here's an excerpt from the pulldown documentation... A B C D D (30fps)
A B B C D You can see that the pictures I got you to post follow this kind of pattern. You only have one clear top/bottom mismatch in each case because animation frames are repeated... the fourth moth picture and the third hand picture are mismatches.

The quickest way to tell pulldown is to look at the frame rate -- the whole purpose of pulldown is to boost the frame rate from 23.976 to 29.97. I got you to post pictures because IMO it was the easiest way to start to explain pulldown... but it sounds like you've sorted it out yourself. Sorry for the extra work! (FWIW, I tend to use the pictures anyway to check for myself, and also to get an idea of how often the telecine pattern changes. If, for example through the entire video it was always the third and fourth frames (in the blocks of five) which had mismatched top and bottom field, then I would know that the telecine pattern didn't change at all... and I could simply use the method given in the AVISynth Pulldown documentation. This may be considered a bit paranoid... but I prefer doing things 'with your bare hands' where possible. But you're absolutely fine sticking with the filters you are using.)

Edit: S-video connector... circle with four pins. If you throw S-video into a Google image search, you get a very satisfying number of pictures.

[HyperSpad]

I just posted sort of a quick response before. Now that I've done some testing I can better respond to what you said, Chainmax.

I took your suggestion with Decimate() and now I run that after Telecide. While I don't see any noticeable advantages of using that for now, I'm sure it'll help with compressing the file and squeezing more quality out of a smaller file size later on. It doesn't harm the motion in the video or anything so I have no problem using it.

Now I'm gonna have to disagree with your suggestion to use post processing. I have thus far found no problems with "Telecided" material that has not been post processed. However, when I tried using post processing, I started getting problems with gaps in the black outlines very similar to the screen shot I posted above from "ELA-2: TDeint(order=1, type=3, sharp=true)"



Alright, I wasn't sure if DeDot() was supposed to be run before interlaced content and telecined content, OR if it was just interlaced content it could be run before. So now I run DeDot() before Telecide and it seems to be doing a bit better. I was a little concerned that it was causing Telecide to not perform as good as it could (I thought I saw some little artifacts on telecined areas when running DeDot first followed by Telecide), but what point is there in using DeDot at all if it's hardly gonna work, so I'll run DeDot first from now on, like you said.

[HyperSpad]

Chainmax:
Looks like you were right about the post processing, I finally found an "interlaced" part that got through. Looks like I'll have to either experiment a bit more with the post processing settings or possibly find another filter to take care of this (I just don't want to end up with gaps in the black outline again).


mg262:
Good to hear I've got something right. Now if only getting rid of dotcrawl was this easy. And yeah, there's definetly something to be said for the thrill of experimenting with filters on your own.

[mg262]

Something else I thought of...it may be that you are getting such bad dot crawl not just because of your connection, but also because of your capture card (for example because it uses a notch filter rather than a comb filter to separate chroma and luma); so it may be that you can improve the situation relatively cheaply by upgrading to another capture card. (There are plenty of cheap good capture cards; IIRC the expensive ones tend to have useless features like capturing straight to MPEG 4 or something like that rather than actually producing better picture quality when used properly.) But this is all something you should ask on the capturing forum if you want to chase it up.

[Revgen]

For anybody who uses TDeint in cartoons:

Make sure to set weave=true somewhere among the parameters. TDint will then use weave if MoComp gives an undesired result. IMHO Weave works better than MoComp in static areas of cartoons. Your encode will take longer but will look better.

[hartford]

Hate to say it, but it's just my opinion.

I've done captures of S-Cry-Ed from Cartoon Network. I did the "telecine, deinterlace" thing and the result was awful. I now just clean it and keep it as
interlaced and encode it as interlaced.

Why was it awful? Because of the "defects" that you posted. Teen Titans and
S-Cry-Ed have some "funky" conversion to NTSC that, when you try to revert to
progressive, causes the result to become jerky.

Just my opinion.

[Chainmax]

I don't know about Teen Titans, but I heard that Scryed is VFR, so neither IVTC nor Deinterlacing will look good. You have to make a variable framerate encode (using Decomb+YATTA or TIVTC) and mux audio and video as an MKV file.

[mg262]

How can one broadcast VFR?

[DarkNite]

VFR is fairly common in cartoons and anime. It's a cost cutting method of production. Why make the animators it took you so long to get locked into that little sweatshop called a studio work so hard making 24 frames for a single second when 8, 12, or 15 will do? Why take the time to render your CGI at 24p and then have to telecine everything when you could just render it at 60hz and edit in video? Just do whatever's easiest, cuts out the most production work, and slap it all out at 30i (here in NTSC land). The low budget studio motto. Besides, all those kiddies watching it on the tube will never know the difference. :sly:

/me points at The Simpsons, Transformers, and half of his anime captures

[mg262]

Thank you. When I read VFR I conjured up something more complex than 12fps animation... I have the vague impression that some of the newer containers or codecs have some arcane 'VFR' features and I was thinking of something along those lines.

Why take the time to render your CGI at 24p and then have to telecine everything when you could just render it at 60hz and edit in video? Just do whatever's easiest, cuts out the most production work, and slap it all out at 30i (here in NTSC land).That answers another question I've been wanting to ask for a long time... although since the CGI computer time is expensive (perhaps less so these days?), and also the 24p material has to be telecined anyway, it's still not clear to me why it's an easier approach.