Output file has odd cadence

[laserfan]

Hi, when I use AGK to convert a 1080i HD transport stream to Xvid, 1280x720 the resulting file appears to duplicate frame 4 at frame 5, i.e. if you count frames (stepping-thru the Xvid file using VdubMod) 1-2-3-4-5, frames 4 and 5 look the same, as if 5 is just 4 duplicated. Then the next frame shows movement again until reaching the next 5th frame. In motion (29.97fps) it doesn't look bad, but I wonder what would cause this?

AGK analyzes the input video, and then deinterlaces it thusly:

[2/8/2006 10:27:45 PM] Source has percentage of interlacing in motion areas: 100.00
[2/8/2006 10:27:45 PM] Source has percentage of telecined patterns: 0.00
[2/8/2006 10:27:45 PM] Source has percentage of progressive patterns: 0.00
[2/8/2006 10:27:45 PM] Source has percentage of interlaced patterns: 100.00
[2/8/2006 10:27:45 PM] Source is considered to be interlaced.
...
KernelDeInt(movie,order=1,sharp=true)

Wondering if anyone can shed some light for me re: the duped frames? I'm assuming the output video would appear smoother w/o the dupes...

[manono]

Hi-

I think I know what causes it laserfan, but I'd need to look at an unprocessed sample to be sure. It sounds like the true framerate should be 23.976fps. Have you tried to go into the Hidden Options to Force Normal IVTC? If you can upload a sample somewhere, I'd be glad to have a look.

[laserfan]

...It sounds like the true framerate should be 23.976fps. Have you tried to go into the Hidden Options to Force Normal IVTC?But CTRL-F9 will only "force ivtc" if the input is deemed to be a hybrid, or so I thought?

The source is NBC's ER on Thursday nights. I don't know if it's shot on film or not. Anyway I will look into how I might make a sample available to you.

Mahalo!

p.s. I will also try AGK Tweaker to IVTC a piece of this. Oh crap! I stupidly deleted the original already! Wait, I do have an earlier episode of ER in its original transport stream form (I use MyHD MDP-130 to capture). I will look at it, verify it does the same thing, and try IVTC'ing.

[laserfan]

I think I know what causes it...It sounds like the true framerate should be 23.976fps...manono I used AGK Tweaker and substituted

Telecide(movie,order=0,guide=1,post=2).Decimate(mode=0)

for AGK's automatic KernelDeInt command and it worked--the 1-2-3-4-(4) cadence is gone. And Gspot says the new output file is 23.976fps instead of 29.970.

Assuming now your hunch is correct--what is the situation and why does AGK not know to IVTC this?

If you still want to see a sample of the original transport stream I can make it available--just PM me with your email address and I'll use yousendit.com.

Thanks, and anxiously awaiting your thoughts on this! Eureka! ;)

EDIT: All my ER conversions-to-Xvid that I've already done (eight or so) show this "problem" of the extra frame--how might I go about fixing them if I wanted to?

[manono]

Hi-

But CTRL-F9 will only "force ivtc" if the input is deemed to be a hybrid, or so I thought?

Yeah, you could be right. I figured that it would force an IVTC under any circumstance.

Assuming now your hunch is correct--what is the situation and why does AGK not know to IVTC this?

You know what regular old telecine looks like, right? 3 progressive frames followed by 2 interlaced frames in every 5 frame cycle. AutoGK can spot that and then apply IVTC. I've seen a couple of examples on DVD where the conversion to DVD was done by adding a complete duplicate progressive frame in every 4 to go from 23.976fps to 29.97fps. This makes for kind of choppy playback, as you discovered with your deinterlaced version, because of the one dupe in every 5 frames. That would be seen by AutoGK as 100% progressive. But yours was seen as 100% interlaced. So, I figured that after adding the duplicate frames they then shifted one of the fields over by one frame to make it 100% interlaced. So, instead of having top field/bottom field going like this:

ABCDD
abcdd

it went like this:

ABCDD
bcdde

I guess that would give 20% progressive frames, so maybe my theory wasn't exactly correct, but something similar must have happened for the IVTC to have worked for you. You can figure it out yourself by opening one of the vobs in DGIndex, finding a section with movement and, using the [ and ] buttons to isolate it, make a D2V project file with Honor Pulldown Flags set, and then open that D2V in GKnot to have a look.

So why couldn't AutoGK spot it? That situation is rare (I've never seen it on a DVD), and if it thinks it's 100% interlaced, it can't do anything but deinterlace it. Sometimes your eyes are the only thing that can determine what you have and how it should be treated.

[foxyshadis]

You could probably add Decimate to ffdshow's avisynth filter section if you want a temporary workaround (for the old ones). Otherwise, you'll probably need to load them into avisynth, deblock and decimate, then reencode, or rerip from scratch. It's odd that it gets it wrong, but sometimes that's the way it goes.

[laserfan]

Thanks you guys--while I'm trying to absorb what you have said let me re-state to you that this is not from a DVD. It is a 1080i OTA transport stream capture of the TV show ER, using a MyHD MDP-130. This board does not "do anything" with the broadcast except get it onto the hard disk.

[manono]

Oh yeah, I shouldn't have said open one of the vobs in DGIndex, but open one of the TS files on DGIndex.

I think that getting rid of the dupes from the earlier attempts will require reencoding. You'll have to add the Decomb filter, with Decimate() (but not Telecide) somewhere in the .avs. I don't know if you can do this in AutoGK or not. You'll be changing the frame count.

foxyshadis' first suggestion was for getting the earlier duped ones to play without the dupe frames. It was a playback suggestion.

[laserfan]

Thanks very much for your patience in responding to me. You set me on the right track re: forcing IVTC, and so long as I can make AGK do this by using the AGK Tweaker (tweakus interruptus), and it makes a nice 24fps XviD file, then I am happy. I will do this on all future conversions of "ER" episodes (wondering to myself if all NBC 1080i programs are encoded in this way).

I don't know if I will try to re-encode any of the earlier conversions or not, they don't look bad, but it's nice to know I have some good files with which to learn DGIndex and Decomb, if we ever again get a rainy day here in TX! :o