should i "force normal ivtc" when told % of progressive patterns is 5.88?

[AGK_J]

usually i get the resulting "Source SEEMS to be pure FILM" in which case i leave it be...
but sometimes i get "Source is CONSIDERED to be FILM", should i "Force normal IVTC when source was detected as hybrid"?
although it's not a hybrid but since it tells me a different result along with something like
"[12/31/2005 2:07:59 AM] Analyzing source.
[12/31/2005 2:11:39 AM] Source has percentage of interlacing in motion areas: 35.09
[12/31/2005 2:11:39 AM] Source has percentage of telecined patterns: 94.12
[12/31/2005 2:11:39 AM] Source has percentage of progressive patterns: 5.88
[12/31/2005 2:11:39 AM] Source has percentage of interlaced patterns: 0.00
[12/31/2005 2:11:39 AM] Source is considered to be FILM."
which concerned me because in results where "Source has percentage of progressive patterns" is over 6.00 it results in a hybrid(and 5.88 is quite close)...so is it just safer to force normal IVTC anyways? or simply let autogk do its job as long as its not detected as hybrid?

...i've never seen any results other then the previous mentioned and "Source is considered to be hybrid (mostly FILM)" in which case i'm sure to use "force normal IVTC" i assume...but just in case, what result would require me to "force deinterlacing"?

PS: what problems could occur if i do not force normal IVTC when source is detected as hybrid? sorry for all the questions, but only 1 of my 19 eps of my alias season 2 rips was detected as hybrid so i had to ask :P

[manono]

Hi-

In the example you gave, it's going to do a normal IVTC (or use Force Film in DGIndex, not sure which), so forcing it won't change anything.

In cases where it's found to be a hybrid and it puts on the special Decimate setting, if it turns out to be really film, that special setting isn't used. It comes into play when you have mostly film, with some really interlaced material. It is possible for it to be detected incorrectly as a hybrid, but you still shouldn't get messed up output.

So, my advice is, unless you've examined it pretty thoroughly yourself and know what you're doing, to leave AutoGK alone.

[laserfan]

I just tried a program (TV capture of a "Masterpiece Theater") that was analyzed thusly:

[1/24/2006 11:45:43 AM] Source has percentage of interlacing in motion areas: 55.37
[1/24/2006 11:45:43 AM] Source has percentage of telecined patterns: 76.64
[1/24/2006 11:45:43 AM] Source has percentage of progressive patterns: 1.81
[1/24/2006 11:45:43 AM] Source has percentage of interlaced patterns: 21.55
[1/24/2006 11:45:43 AM] Source is considered to be hybrid (mostly FILM).
[1/24/2006 11:45:43 AM] Looking for optimal hybrid thresholds.
[1/24/2006 11:46:21 AM] Found threshold of: 1.97

I let it run in auto where is does the Decimate thing, and I also tried a Force IVTC on it, and in both cases it still plays herky-jerky. Am I out of options w/AGK?

[manono]

Hi-

That's one of the worst kinds to have, as far as trying to get it to play smoothly at a single framerate. The Decomb hybrid modes are far from perfect, and even if it was analyzed correctly, there's just no way that the video parts are going to wind up playing as well as the film parts. But no, you're not out of options. They are, as I see them:
1. It's a hybrid and Decomb did the best it could. Live with it if you prefer to work only in AutoGK.
2. It's a hybrid, but you think you may be able to do better yourself. Use GKnot, edit the .avs, and try SmartDecimate (or maybe TIVTC's hybrid modes) as the IVTC. It was developed with hybrids in mind and you may (or may not) like the results better.
3. AutoGK didn't analyze it correctly, it's really pure interlaced video, and IVTC was set incorrectly. If that's the case, I'd recommend going into the Hidden Options and "Force Deinterlace". Of course, it'll help if you know how to analyze it yourself. Sometimes nothing beats the eyes for this kind of work. Make a D2V with "Honor Pulldown Flags", open it in GKnot, find a part with motion, and look for interlacing in every frame. If you find it in many different sections of the video, then you'll know it's interlaced and should just be deinterlaced, with no IVTC applied. Use Force Deinterlace in the Hidden Options of AutoGK. If you're not up to the analyze task yourself, I'd recommend doing that next to see if it plays more smoothly. Because it'll have 25% more frames (29.97fps .vs 23.976fps), for the same quality you should allow for either a larger file size, or a lower resolution.
4. You have something really screwy, such as field order problems (common with captures), or some kind of blended video, which will require some serious analyzing and AviSynth skills. AutoGK may not be the best program for this, and it's time to read up on how to detect field order (Run the Preview in DGIndex is one way), and what to do about it, and how to detect blends. Put on a smart bobber such as KernelBob to check for blends. Let's hope it's not #4.

[laserfan]

...I'd recommend going into the Hidden Options and "Force Deinterlace"...That's trivially easy to try--I will give that a go.

And thanks for the thoughtful suggestions; I have put-off learning how to do analysis for a long time now. Maybe this will be a good test case for me ("you only truly learn something new by solving a problem that is meaningful to you").

But one more thing I know to try first (since you mentioned that Bob character!) is to run it thru Fairuse Wizard which I think uses some different techniques for both analysis and deinterlacing/bobbing.

Mahalo!

[laserfan]

Hmmm, I just thought of another option. It is possible if not likely that the body of this program is film-based, and it is only the opening title, Masterpiece Theater boilerplate, and closing boilerplate that are video. Maybe I could cut the video-up, process each piece differently, and then join them (not sure Vdubmod can do this for Xvid files but it's gotta be possible?).

The lazy man's solution perhaps.

[manono]

Sure, you could always try a different program to see if it does any better.

Bobbing doubles the framerate by turning each field into a frame. It's a good way to check for blends. If you see the ghosting/double images in the bobbed fields, then that's the proof of blending. I don't really think your problem is blending, though, and FairUse won't be bobbing anything, as far as I know (not having used it for at least 5 years).

Yes, it's not uncommon for the main body of a show to be film, and the opening and ending credits be video. I have no idea if it's true in your case, but, if I'm reading that log correctly, I don't think it's pointing to that possiblility. You say the main part of the show is playing jerky, right? So that seems to indicate that it's not film based.

[laserfan]

...Bobbing doubles the framerate by turning each field into a frame...Thanks manono. I ran Fairuse Wizard on the program--it suggested treating the program as "Native" vs IVTC but it still looked bad when I encoded a piece that way. Then I selected also its "Bob" option w/Native and it all looks good, both the boilerplate titles and the body of the program.

FUW doesn't have a Help file or otherwise explain exactly what these options do or how it does them. I guess I am a bit surprised though that for as sophisticated and complete a solution AGK is for doing these conversions, that no easy option exists to perform a similar function (bobbing).

Maybe I will try AGK Tweaker some time to see if I can find a path--using FUW for this was a PIA cuz I had to: save the mpg as Elementary streams, mux them again using MuxMan to make a VIDEO_TS titleset, use DVD Shrink to make an ISO from the titleset, then run FUW on the ISO (cuz unlike AGK it only works on DVDs or ISO files). Whew!

Anyway thanks again, and my apologies to AGK_J for hijacking his thread! :o

[manono]

Hi-

Then I selected also its "Bob" option w/Native and it all looks good,...

Interesting. So I guess it bobs it and then selects one of the bobbed fields, perhaps something like:

KernelBob(Order=1).SelectEven()

I don't know how good its bobber is, but you might find some thin diagonal lines and look for evidence of stairstepping/aliasing/jaggies. Electric lines are always a good thing to look for. I'm not sure why the Bob/Native option works better than the Native option (which I assume is a deinterlace) though. In any event, if it worked for you, then good. What's the framerate, by the way? Is it still 29.97fps, or is it really a straight bob resulting in 59.94fps?