AutoGK Analysis, how good is it?

[hotwings]

Ok i've done some searches of the forums but couldn't quite find what I was looking for so bear with me, i'm hoping some of your resident gurus can help me out.

Now I use autogk every now and then, a lot of times to help determine if my dvd source is NTSC FILM or VIDEO. It seems to work ok most of the time, but the other day it missed a FILM source with the analysis saying that the movie was 99% progressive with 0% telecined frames. It left me quite baffled how it could be so wrong. (to determine that it was wrong i opened up the output avi file in virtualdub and sure enough every 4th and 5th frame was a duplicate).

I came to the conclusion after some investigation that the reason the analysis had trouble with the dvd was that it was HARD ENCODED to NTSC VIDEO even though it was obviously shot in FILM, thus since it didn't use any 3:2 pulldown flags autogk had a problem detecting it?

I guess what I'm getting at is what is THE BEST way to determine if a source truly is FILM or VIDEO? The eyeball method seems to work, but there have been other times where autogk has determined a movie to be FILM. I open the VOBS in virtualddub and sure enough the framerate is about 24fps. So i encode it without telecining at 29.97 fps just for kicks, and yet when i eyeball the resulting output file, I can't spot any duplicate frames. Guess i'm a little confused.

EDITED TO ADD: Just how good is the eyeball method btw? If i understand telecining, sometimes half of two different frames are combined to form a new frame for 3:2 pulldown purposes. Is it possible to eyeball and miss duplicate frames because the new frame created really is different enough not to look like a duplicate, but rather a unique frame?

[manono]

Hi-

AutoGK isn't perfect with its analysis, but it's better than any other automatic method, and although it's not better than the eyes of someone that knows what he's doing, it is better than you.

Now I use autogk every now and then, a lot of times to help determine if my dvd source is NTSC FILM or VIDEO.

Why use AutoGK for that? DGIndex can tell you that in a second? However, in DGIndex and AutoGK terms, there's no such thing as NTSC FILM. NTSC is by definition 29.97fps, and therefore not Film.

It seems to work ok most of the time, but the other day it missed a FILM source with the analysis saying that the movie was 99% progressive with 0% telecined frames.

How do you know it wasn't on the DVD like that, progressive 29.97fps with every 5th frame a duplicate frame? Did you check out the source VOB? If it was progrssive 29.97fps with every 5th frame a dupe (very rare, but not unheard of), then, by definition, it's not Film and it wasn't telecined.

It left me quite baffled how it could be so wrong. (to determine that it was wrong i opened up the output avi file in virtualdub and sure enough every 4th and 5th frame was a duplicate).

You haven't proven anything. But AutoGK can't spot that kind of thing and pull out the dupes, assuming the source VOB was like that as well. Got a VOB sample? During the analysis, when it looks for evidence of telecining, it's looking for 40% of the frames to be interlaced (2 of every 5 frames that show movement). It has no way to spot dupe frames. But, as I said, that kind of thing is very rare.

I guess what I'm getting at is what is THE BEST way to determine if a source truly is FILM or VIDEO?

The DGIndex Preview. But then, that's not what you're really asking. You are including hard telecine in your definition of Film, but as it's really already telecined and encoded as interlaced 29.97fps, it's really Video. It can be IVTC'd back to 23.976fps (something AutoGK is good at spotting and doing), whereas if it was shot using video cameras, and all frames are interlaced, it can't be IVTC'd, but only deinterlaced (something AutoGK is good at spotting and doing). What's the best way to tell the difference between the 2? Make the D2V using Honor Pulldown Flags, and the .avs with no filtering. Open it in VDub(Mod), find a place with movement every frame and look at the interlacing pattern. If every frame is interlaced, it can't be IVTC'd. If, in every 5 frames, 3 are clean and progressive and 2 are interlaced, it can and should be IVTC'd. If it shows as film in DGIndex, then you make the D2V using Force Film. If it shows as Progressive NTSC in DGIndex, as in your earlier example, have a look at the frames. If every frame is different and unique, leave it at 29.97fps. If every 5th frame is a dupe, use a decimator (Decomb's Decimate(), or TIVTC's TDecimate(), for example) to remove the dupes.

I open the VOBS in virtualddub...

VDub doesn't open VOBs. I take it you mean VDubMod.

but there have been other times where autogk has determined a movie to be FILM. I open the VOBS in virtualddub and sure enough the framerate is about 24fps. So i encode it without telecining at 29.97 fps just for kicks, and yet when i eyeball the resulting output file, I can't spot any duplicate frames. Guess i'm a little confused.

Yep, you're confused.

Just how good is the eyeball method btw?

If you know what you're doing, it's the best way, but what followed in that paragraph didn't make much sense to me.

[hotwings]

honestly dude, when did i say i knew everything? i asked for some help because i wasn't sure and instead i get some defensive douchebag response, almost as if you wrote autogk yourself. if you did that's your problem. I figured out what i needed to before your response, and you attacking me for stupid semantics like NTSC Film instead of getting to the root of the question is beyond retarded really. You would have thought i insulted your mother. i asked for help from resident gurus not resident pricks. anyone who is too into himself and thinks he knows it all to be so brash has to be one ginormous tool...and precisely the kind of person who is too proud to learn from others. i could do without your kind of 'help'.

[manono]

There wasn't anything defensive in my response. If you can't take some constructive criticism, then what are you doing asking ignorant questions maligning the analysis portion of a very good program? There's nothing wrong with being ignorant, mind you, but if you don't or won't learn from the response, then that goes beyond ignorant and crosses over into stupid. You made a number of incorrect statements in your post, and I took quite a bit of time to try and set you straight.

I figured out what i needed to before your response, and you attacking me for stupid semantics like NTSC Film instead of getting to the root of the question is beyond retarded really.

Sorry, but it's much more than semantics, and goes to the root of your misunderstanding of what's going on. I'll ask again; do you have a VOB sample to prove your contention that AutoGK was wrong when you concluded that "it missed a FILM source with the analysis saying that the movie was 99% progressive with 0% telecined frames."? If you don't have anything intelligent to say in response, then I'll retire and let someone else take over.

[hotwings]

yes i easily admit that by including hard telecining i was mistaken, nothing wrong with that. I admit i am wrong all the time. You on the other hand have to go the extra mile of degrading someone's intelligence because hmm their life experience doesn't mirror yours. I could find a dozen things i'm good at that you might sound somewhat buffoonish when discussing a subject, difference is that I would point out the mistakes without the extracurricular bs. this thread has outlived its usefulness..thanks but no thanks I really don't feel like exploring this any further. Like i said this 'idiot' figured it out all by himself..so you can move along now.