How to become an encoder?

[Skyward]

I'm sorry, I sort of generalized when I said "Interlacing". While Telecine (specifically a 3:2 pulldown and its varients) is the most commonly used way to convert FILM->NTSC, its not the only method there is. There is also a Telecine pattern (2^12:3) used to convert FILM->PAL that does not involve speeding up the audio.

[Mentar]

Originally Posted by jfs

No, the FILM to NTSC conversion does is NOT "just through interlacing". The process has a very specific name, it's called TeleCine (convert into TELEvision from CINEma) and it merely exploits the field-based nature of video.

Almost. TeleCine is generally correct, but it's something else than "exploiting the field-based nature of video". Obviously the source is 23.976 frames per second _progressive_ full-frame footage (called FILM). However, the format of NTSC DVDs requires 29.970 frames per second _interlaced_.

So, the required process isn't just to interlace the full frames, but also to raise the framerate by 6 frames per seconds. This process is called "3:2 pulldown" and done by introducing duplicates into the sequence of half-frames (fields).

While it does result in interlacing artifacts when played back on progressive scan equipment without using a proper IVTC (InVerse TeleCine) pulldown filtering it's still not true interlacing.

Yep. IVTC is the inverse process of 3:2 pulldown, it tries to reconstruct 23.976 progressive footage out of 29.970 interlaced.

Edit: Oh, and I forgot to say: NTSC->PAL conversions can never be saved fully. NTSC is, as mentioned, 29.97 (well, 30/1.001) fps field based video, while PAL is 25 fps field based video. Any conversion from NTSC to PAL will lose video frames or fields through some kind of irreversible process, unless you elect to slow down the video by about 17 percent, which will be very noticeable, and certainly not desirable. If the original material is FILM, what is usually done is that it's instead sped up from 24 fps to 25 fps, producing progressive field-based video. This speed up is not too noticeable and is standard practice on most PAL land video/DVD releases of cinema film productions.

Right. That's the theory at least. Unfortunately, all too often, PAL DVDs are sucky and dirty conversions not from the original FILM 23.976 progressive material, but rather from 29.970 interlaced. And the results are pretty ugly.

[checkers]

Mostly the anime PAL I see is field blended. Sometimes fields are duplicated instead of blended. In this case, it doesn't matter if the source was 24 or 29.

[shyrin]

I'm fighting with such a piece of crap right now. Its really bad you have to face three problems: Ghosting, judder and bad flickering...

actually I wanted to ask another question (that might be stupid):
Is there anything Yatta can do what you can't do with avisynth? I mean should I learn to use Yatta or am I fine just with avisynth?

[martino]

Originally Posted by shyrin

actually I wanted to ask another question (that might be stupid):
Is there anything Yatta can do what you can't do with avisynth? I mean should I learn to use Yatta or am I fine just with avisynth?

Uhm... AviSynth is a frame serving application, and YATTA is a tool for IVTC and video clean up tasks. Somewhat two different things. If you want me to answer your question, then I guess a close answer would be "yes". Certain tasks that YATTA allows you to do are not easily achievable with just AviSynth itself (for example, freezeframing, sectioned filtering and some others - and of course the IVTC part). And it might be a good idea to learn how to use it, however don't count on being 100% sane after the process.

[TheFluff]

Eh. YATTA is just a severely over engineered and highly obfuscated GUI for Avisynth in general and TIVTC/decomb in particular, really. All it does is making it easy to write ridiculously long and complicated Avisynth scripts that you'd never write otherwise (plus save some override files for fieldhint, TIVTC and/or decomb).

Granted, it does have some pretty nifty automated fieldmatching improvement logic (pattern guidance in particular) and tools (bitrate calculator/cropping & resizing etc.) that don't exist in Avisynth, but in terms of what you can "do" to video it has the exact same limits as Avisynth since all it does is produce Avisynth scripts.

[Unearthly]

Originally Posted by martino

And it might be a good idea to learn how to use it

Unfortunately, the manual is very outdated so it's very difficult to learn how to use, no thanks to:

Originally Posted by TheFluff

severely over engineered and highly obfuscated

I tried learning it and gave up, just because I was frustrated at having no idea how to do anything. Instead I've learned how to do stuff natively in avisynth, which for the most part isn't that bad. For example:

Originally Posted by martino

(freezeframing, sectioned filtering and some others - and of course the IVTC part)

FreezeFraming - FreezeFrame

Sectioned Filtering - Not hard with a little helper function I wrote:

function Replace(clip input, clip replace, int startFrame, int endFrame)
{
	first = input.Trim(0,startFrame-1)
	second = input.Trim(endFrame+1,input.framecount-1)
	return startFrame == 0 ? (replace+second):\
	(first+replace+second)
}

Example:

filtered =  Trim(500,1000).FFT3DGPU()
Replace(filtered,500,1000)

And IVTC isn't that bad with at least the few modern DVDs I've worked with, which required < 10 manual override sections per episode (usually just fades). From what I've heard older DVDs and PAL DVDs are different beasts that could benefit a lot from an automated tool, though.

[jfs]

Unearthly, correct that freezeframing and sectioned filtering *by themselves* aren't very complicated, but the hard part of it is picking the right frame numbers to do it on, and that's where YATTA comes in handy, since it's a graphical interface for it: find the right frames and press the right keys, instead of zooming around in VDub, switching to Notepad or whatever and then entering the frame numbers manually. It's not hard, it's just much more tedious.