IVTC - could someone help?

[Gew]

So what will one do at such ugly moment? (my question again), is it possible to do for instance Yadif (for deinterlacing) and TFM().TDecimate() for IVTC...?

[poisondeathray]

There is no good single answer, because it it depends on the source and what exactly it is, and how bad it is...

For most decent sources that are primarily telecined (maybe a few sections have breaks) , TFM().TDecimate() should work. This should work for most recent Hollywood DVD's for example.

[Gew]

Oki!

Now say I've got a movie as PAL DVD á 25 fps, which I wanna bring back to 23.976 fps..? TFM().TDecimate() gives me 20 fps in this scenario. I've read info/wikis on TFM() aswell as TDecimate() but they seem überadvanced, more or less like Greek to me. Doing ChangeFPS(23.976) or ConvertFPS(23.976) here would be cheating too, right? So what's the right line?

[poisondeathray]

Nope, it depends how the PAL DVD was made, but generally what you want to do is a "PAL slowdown." There should be several guides and steps how to do that if you search, the audio has to be slowed down as well

AssumeFPS(24000,1001)

And do the audio in eac3to or besweet

AssumeFPS just does a simple slowdown (no change in frame count) , so if the PAL DVD was made from a simple speedup from 23.976, that should reverse it

[Gew]

Oh, I see.

Then I must ask (even though it may or may not be very a much logical subsequent fact to most ppl.. By doing this slowdown, would I be tampering with the original film (~23.976), I mean by doing the slowdown of video and audio, or would I actually brining audio and video back to its original? In other words, are DVD movies at 25 fps generally pitched up from their hollywood studios original video framerate/audio speed? Or does the studios that originally do the film->25fps DVD release process do this in some advanced manner (some sort of IVTC, I could imagine) that preserves fps & audio speed?

I guess (TLDR-isch) my real question is if I'd -- generally -- be sabotaging the fps/audio speed of the original film, or would I be restoring/fixing it, by doing AssumeFPS?


Ty in adv.
Regards~

PS. Is AssumeFPS(24000,1001) more correct than AssumeFPS(23.976)..?

[poisondeathray]

I think that was already answered. It depends on how the PAL conversion was done. Most do a simple speed change. To reverse it you reverse the speed change

24000/1000 is more correct , 23.976 is rounded to 3 decimal places and can cause gradual sync error

[Gew]

Ok, sorry, my bad.

And just to be clear:

AssumeFPS(24000,1001) = 24000/1000 ?

[poisondeathray]

yes

[manono]

AssumeFPS(24000,1001) is correct. It's about the same as 23.976fps.

Edit: Apparently I posted right after you did, poisondeathray. How can AssumeFPS(24000,1001) = 24000/1000 ?

[Gew]

Thanks!

Also, now I happen to have an example on the scenario that I tried imaginging earlier in the thread.
It's NTSC DVD of The Final Break (round-up episode on that prison series).

Now, this is what I did.

First, I ran a util that I've had _much_ use for, called PGCDemux. I pulled out .m2v & .ac3 with this.
Second, I opened the .m2v in DGIndex and simply rendered the project as .d2v.

Now, that info-window to the left is showing "Video Type: Film".

Great. I do azid.exe -a audio.ac3 a.wav, just to have pure PCM audio ready.

Now, avisynth script. I do AudioDub(mpeg2source...,wavsource....), and add TFM().TDecimate().

The script is playing fine, even due its hasnt much confidence ('cuz I don't have a trained eye for this stuff) my first look showed no sight of dupe frames, dropped frames, or anything. The video played smoothly and showed 23.976 in Media Player Classic media properties.

However, as I paused and "stepped", I could easily see that the video was interlaced. Very visible horizontal fields. I then added Yadif(order=0, mode=0) just below the AudioDub().

The video looks fine now. But am I doing it right?
Or should I have placed the deinterlacing _after_ TFM().TDecimate()..?


Ty in adv.
Regards~

[poisondeathray]

Edit: Apparently I posted right after you did, poisondeathray. How can AssumeFPS(24000,1001) = 24000/1000 ?

I was referring to the 23.976 decimal appoximation vs. 24000/1001 fraction comparison posted earlier. The syntax for AssumeFPS as given above for avisynth is correct , I should have been more clear, but I think Gew got the jist of it

Or should I have placed the deinterlacing _after_ TFM().TDecimate()..?

Definitely not. You should never combine full IVTC and deinterlace after.

Gew - I don't know, it might be that your source is interlaced, I don't have ESP or a crystal ball. Did you got through the steps in the guide?

If it's interlaced => deinterlace it
If it's telecined => reverse telecine it (IVTC)
If it's PAL speedup => do the reverse (ie. PAL slowdown)
If it's up => go down
etc...

It's just reversing everything

Maybe post a sample of the .m2v if you want confirmation. Sometimes you need other post processing to make it better if the IVTC isn't perfect. If DGIndex reported 100% film then it should have worked fine even in "forced film" mode when making the d2v

[Gew]

Yeah. I figure. So, never do deinterlace after IVTC.
Thus I assume, in my example, since it is indeed film, I shall then use ie. SmoothDeinterlace().TFM().TDecimate().

[poisondeathray]

Yeah. I figure. So, never do deinterlace after IVTC.
Thus I assume, in my example, since it is indeed film, I shall then use ie. SmoothDeinterlace().TFM().TDecimate().

I don't think that's a good idea either. My point was not to combine both. By other post processing I mean vinverse() or other filters to get rid of residual combing

If you deinterlace first, your fields will be gone, so there IVTC or field matcher will be useless

Not sure why you like smoothdeinterlace, pretty crappy quality. Even the author says so.

Post a sample

[Gew]

Hmm. Strange thing just happened. I ran the .avs with _only_ TFM().TDecimate(). again, and now, I couldn't see any fields. I must have forgotten one quicksave or something earlier. So, with this said, I suppose that this is one of these times when TFM().TDecimate(). is how to best proceed. Here's a sample from the original .m2v (cut out with Avidemux).

http://carl.1gbps.se/prbr.mpg

So, how's it hanging?

[poisondeathray]

Yes, standard 2:3 telecine

If you go through that mini-guide, it is very simple to figure out ; basically when you separate the fields: 2 repeats, 3 repeats, 2 repeats, 3 repeats etc...

[jagabo]

Hmm. Strange thing just happened. I ran the .avs with _only_ TFM().TDecimate(). again, and now, I couldn't see any fields. I must have forgotten one quicksave or something earlier. So, with this said, I suppose that this is one of these times when TFM().TDecimate(). is how to best proceed. Here's a sample from the original .m2v (cut out with Avidemux).

http://carl.1gbps.se/prbr.mpg

So, how's it hanging?

That clip uses pulldown flags, it's not hard telecined. It doesn't need TFM().TDecimate(). Just use DgIndex in Forced Film mode (unless other parts of the video are hard telecined).

[Gew]

Nice. I've been reading on about the 3:2 pulldown now. I'm not very good at math, but just to look for schematics wouldn't be too much of a hazzle, I suppose. Only, where shall I look for that aaabbcccdd... thingie? Do I have to set my .avs script to act like histogram() or something?

[manono]

Only, where shall I look for that aaabbcccdd... thingie?

In the case of that sample (which may or may not be representative of the entire video), if you run the Preview (File->Preview) it says Film/Progressive. It's 100% film and you don't need to use Honor Pulldown Flags and apply an IVTC in the script. You set the Field Operation to Forced Film and create a progressive 23.976fps D2V project file. Use that in the script with no IVTC or deinterlacing or anything.

When the Preview says NTSC/Interlaced then you make the D2V using Honor Pulldown Flags and open an unfiltered script in VDub(Mod). If you see 3 progressive and 2 interlaced frames in every 5 frame cycle, it needs an IVTC. Or, put on SeparateFields() or a bobber (Yadif(Order=1,Mode=1)) and do the same thing. Then if you see the aaabbcccdd, you should also apply an IVTC. If, when doing the first test every frame is interlaced, it's not a candidate for IVTC. If, when doing the second test there are no duplicates (it goes abcdefg...), it's not a candidate for IVTC.

[Gew]

Wow. Look. I've never attempted this before, 'cuz -- tbh -- when having read posts about that famous "3:2 pulldown" I just gazed and often it seemed like further arithmetics was to come, and like I said, I suck at math. Therefor I'd decided that that thing was above my head. Anyways, now, I when being given so straight diretions, I decided to give it a try. At first, I just bumped my .avs into VDubMod and zoomed in 300% and started nudging arrow right. Didn't see one single interlaced frame. Aawergh. Then it struck me, I had changed to "Force film" setting in DGIndex and (re)saved the .d2v after that, perhaps it had something to do with that. I changed back to "Honor pulldown flags" and re-saved .d2v again. Re-loaded .avs in VirtualDubMod. Now, it made sense. I could -- honest to god -- say that I catched and understood how -- in this case -- I got 2 interlaced fields, then 3 normal ones, 2 interlaced, 3 normal ones, I sat for a couple of minutes just nudging right arrow and kept counting. It was such a treat. Now, with this little thing in knowledge I feel I could expand in the space of videometrics.

So, have I understood you guys correctly. If I get video material that I get this "3:2 behaviour" in VdubMod with (using Video / Field Operations / Honor pulldown flags in the saved DGIndex .d2v), it means that I do not need to do TFM().TDecimate(), but just change to Field Operations --> Force Film and save .d2v and that's that, correct?

Then, I can't help to ask myself, what's with TFM().TDecimate()..? Is this to use if I cannot see a straight pattern (ie 3:2) when "pre-checking", sort of the semi-best alternative, where that Force film is the best?

I've learned much to night.
Ty in adv.

Regards~

[manono]

So, have I understood you guys correctly. If I get video material that I get this "3:2 behaviour" in VdubMod with (using Video / Field Operations / Honor pulldown flags in the saved DGIndex .d2v), it means that I do not need to do TFM().TDecimate(), but just change to Field Operations --> Force Film and save .d2v and that's that, correct?

No. Run the Preview. If it's 100% film, use Forced Film and that's it. If it's NTSC then you make the D2V using Honor Pulldown flags, do the test as described earlier, and if it passes the test apply an IVTC. Only apply an IVTC (TFM().TDecimate()) if it's hard telecine - the telecine encoded into the video. It's pointless, much slower, and prone to mistakes to apply an IVTC when there's no need - when it's been soft telecined (shows as Film in the DGIndex Preview).

If you do a Forced Film on hard telecine you'll wind up with an interlaced and jerky-playing mess.

Then, I can't help to ask myself, what's with TFM().TDecimate()?

You IVTC when it's hard telecine, or it's a mix of hard and soft telecine.

[Gew]

If it's NTSC then you make the D2V using Honor Pulldown flags, do the test as described earlier, and if it passes the test apply an IVTC.

And if I get a different, say random pattern, it's not of pulldown attribute, and I will have to eg. just AssumeFPS(24000,1001),Yaidf() it for interlacing and rate restoration ..?

[manono]

Well, if it's a movie there aren't all that many choices. Movies are, by definition, progressive and originally shot at 24fps. So, if you see a different pattern (and I'm not including animations in this, as they can be created at a variety of different framerates), then perhaps it's been field-blended as a result of a poor standards conversion. If you come across something that doesn't fit the pattern, you might post a sample so we can have a look.