A question about an old Fulci thread.

[StrobeLightEpileptic]

I didnt' want to raise a post from nearly 4 years ago, so I thought I'd just post about it. I've been following Fulci's avisynth guide to convert Pal DVD to NTSC:

https://forum.videohelp.com/topic294142.html

I was doing okay until I got to the TMPGENC part. He says:

Using this method you cannot use the WIZARD MODE in TMPGEnc Plus. Instead you will do this the "manual" way first loading the NTSC template. For the source you select non-interlaced and you need to change the FPS to 25fps. You will end up with a high-bred output that has a NTSC resolution of 720x480 but with 25fps which is the PAL format.

Well, I was able to load the template and changed it from interlaced to progressive, but I was unable to change the framerate from 29.97. That option was grayed out.

Right now I'm letting it run at 29.97, only because I don't know what else to do.

How am I supposed to change that?

[SLK001]

I don't know about your template, but if you are trying to encode in progressive mode, the frame rate is implied to be 23.97fps, thus the greying out of the 29.97fps option. Depending on your version of TMPG, there maybe a "Apply 2:3 pulldown flags after encoding" option somewhere.

Or, you could just apply the 2:3 pulldown after everything is encoded using PULLDOWN.EXE.

[StrobeLightEpileptic]

Okay, I read his post further where he said I could put in the ConvertFPS(23.976) and then just use the template wizard in tmpgenc. So, I did that. The resulting file is an m2v and MPEGValidator won't accept it's extension.

So, I tried GSpot and am given inconsistent results: ie; It says Pics/s is 23.976 but Frames per second is 29.97 and the Fields at 59.94 is consistent with 29.97.

It says the video is interlaced.

Now, this was encoded by tmpgenc with "3:2 pulldown on playback" setting. Is it right that GSpot would be calling this interlaced?

Also it looks interlaced in VDub

[AlanHK]

inconsistent results: ie; It says Pics/s is 23.976 but Frames per second is 29.97

That's not inconsistent, that's what 2:3 pulldown does.

[StrobeLightEpileptic]

That's not inconsistent, that's what 2:3 pulldown does.

Well, okay. I went ahead and made a disc and put it in the player. And when I click the info buttons on the remote it does report progressive video.

But, I don't know what I do wrong, because the video is jumpy. The first thing I would think is that the fields were reversed. But the original file was TFF so I listed that in the avisynth script that way.

What other reasons would you get jumpy video.

[manono]

But the original file was TFF so I listed that in the avisynth script that way.

What about in the encoder, also TFF? What's the full script?

[jagabo]

You can unlock any setting in TMPGEnc Plus by right clicking on the text label and selecting Unlock. You can unlock all settings by pressing the Load button on the main interface and selecting the Unlock.mcf template.

[StrobeLightEpileptic]

You can unlock any setting in TMPGEnc Plus by right clicking on the text label and selecting Unlock. You can unlock all settings by pressing the Load button on the main interface and selecting the Unlock.mcf template.

Thanks jagobo. Yes, I ran into this last night when I was using Fulci's tmpgenc/dgindex method.

I'll have to try the avisynth method again now, once I check the quality of the disc with this other method.

[StrobeLightEpileptic]

But the original file was TFF so I listed that in the avisynth script that way.

What about in the encoder, also TFF? What's the full script?

The encoder was tmpgenc. Yes, it reported TFF.

The script was a carbon copy of Fulci's (minus the paths):

LoadPlugin("C:\DGMPGDec\DGDecode.dll")
LoadPlugin("C:\LEAKKERNELDEINT\LeakKernelDeint.dll ")
Mpeg2Source("C:\PAL_MOVIE\VIDEO_TS\mainmovie.d2v")
AssumeTFF()
LeakKernelBob(order=1)
LanczosResize(720,480)
ChangeFPS(23.976)
ConvertToRGB24(interlaced=false)

Since I now know how I can change tmpgenc's default settings, I should re-do it again without the ChangeFPS line. I've never been able to achieve smooth video with any of the avisynth 23.976 settings.

I got close once, but the only way I can get smooth video so far is to leave it interlaced.

[jagabo]

If you were feeding TMPGEnc progressive frames at 23.976 fps TMGPEnc should be set up as follows:

Advanced tab, Video Source Type: Non-interlaced (progressive)
Video tab, Encode Mode: Non-interlace
Video tab, Frame Rate: 23.976 fps (internally 29.97 fps)

Make sure your PAL source is really interlaced as opposed to progressive frames captured out of phase. If it's out of phase you can easily restore the progressive frames with SeparateFields().Trim(1,0).Weave(), or Telecide(), or TFM(). Then encode that at 25 fps progressive and use DgPulldown to add 3:2:3:2:2 pulldown flags.

[StrobeLightEpileptic]

ally 29.97 fps)

Make sure your PAL source is really interlaced as opposed to progressive frames captured out of phase. If it's out of phase you can easily restore the progressive frames with SeparateFields().Trim(1,0).Weave(), or Telecide(), or TFM(). Then encode that at 25 fps progressive and use DgPulldown to add 3:2:3:2:2 pulldown flags.

Yeah, this interlaced vs out-of-phase frames issue really has me hornswoggled. When I looked at the seperated fields in VDub, I couldn't exactly say it was interlaced or progressive, so it crossed my mind that maybe it was the blended frames issue, (am I right in thinking out-of-phase and blended frames are interchangeable terms?).

The disc I am working with is a R2 Pal import. I tried the patch method on it and it works very well on my machine, but not my sister's. So, I'm doing this more out of curiosity than anything else.

But, okay, when I play this in my Apex and use the remote to Display the disc info, it says it is interlaced video. Is that a reliable indicator that it is actually interlaced?

[manono]

ChangeFPS(23.976)
.
.
I got close once, but the only way I can get smooth video so far is to leave it interlaced.

You start with a 25fps interlaced video, deinterlace it, drop one unique frame every second, and then wonder why it's playing jerky? Do you understand what ChangeFPS does? Maybe you meant to use AssumeFPS(23.976), but then you have to slow the audio also. Or maybe you meant to do what FulciLives actually said, which was to leave it at 25fps for encoding, in which case you don't touch the native framerate at all.

If it's a movie you don't deinterlace it. Is it really interlaced? Do you have a small sample so we can see what you really have and give more informed advice about how it should be treated?

am I right in thinking out-of-phase and blended frames are interchangeable terms?.

No. Field-shifted video is progressive and clean once the fields are realigned again. Blended stuff has to be unblended.

Is that a reliable indicator that it is actually interlaced?

No, not really. Just about all NTSC DVDs are interlaced once you take pulldown into account. It might have been encoded as progressive 23.976fps, but once pulldown is applied it then outputs interlaced 29.97fps.

[StrobeLightEpileptic]

If it's a movie you don't deinterlace it. Is it really interlaced? Do you have a small sample so we can see what you really have and give more informed advice about how it should be treated?

Yes, I'll upload something tomorrow. I deleted all the files off my harddrive to make way for other things I wanted to work on. I'll have to rip it again tomorrow, and I'll upload a piece.

The Magnificent Ambersons.

[StrobeLightEpileptic]

Okay, I've uploaded 2 clips. The first clip is a just a motion clip of horse and carriage.

http://rapidshare.com/files/319655584/1st-amb.vob.html
MD5: EC6211BA4BDD352898FB760F349DF229

The second clip has a larger file size.

http://rapidshare.com/files/319655585/2nd-amb.vob.html
MD5: 1533B63FC215B69B542DD8440104E111

I include it for two reasons. One is the shimmering on the door when it is closed. I have noticed some shimmering on the dvd when I watch it on the tv. This example is where it is most obvious, but you can see more subtle instances through out the film. I'm wondering if this is caused just because PAL video is being used by an NTSC player and TV. It doesn't make sense that it would if you can see it on the computer. So, what causes it?

The second reason for the second clip is the unusual fields. I believe that you are watching this in something like virtualdub with the view fields filter. And the fields in this clip show why I am confused by this film, (whether it is interlaced or progressive or blended or out of phase - the last two terms I don't completely understand.) You can see that the fields do not always match up, but you can also see, what I think of as strange artifacting in some of those fields.

[jagabo]

I'm waiting for the downloads but regarding Field shifted PAL:

Standard definition video is always broadcast (or carried over the wire) interlaced -- meaning one field at a time. A progressive film source, sped up from 24 fps to 25 fps, might be broadcast as top field first. Using numbers for the film frame numbers

1 2 3 4...

and t/b for top and bottom fields, this is what comes over the air, and what you see on an interlaced TV:

1t 1b 2t 2b 3t 3b 4t 4b

If a capture card captures that top field first:

1t+1b 2t+2b 3t+3b 4t+4b

each frame of captured video has both the top and bottom fields from the same film frame. They will look progressive.

But if a capture card captures that bottom field first:

1b+2t 2b+3t 3b=4t 4b+5t

each frame contains a field from one film frame and a field from another field frame. Each frame looks interlaced (when there is motion). This is what is meant by out of phase PAL video. (This video will play back properly because the video is flagged bottom field first so the player knows which of the two fields of each frame to display first.) These can be easily converted back to progressive frame by recombining the field pairs.

Got the files now.

Your videos are field blended NTSC to PAL conversions. Rather than speeding the film up to 25 fps, or repeating every 24th frame to get 25 frames, some fields are blended together from adjacent frames of film to create something that looks a little smoother on interlaced TVs.

Think of it something like this. Suppose we have film frames that are supposed to be displayed for 8 units of time:

time --> 
1-------2-------3...

So the 1 followed by 7 dashes indicate an 8 unit duration of time that frame 1 is displayed. But you record that with a camera that captures frames (or fields) that only last 3 units of time:

time --> 
1-------2-------3...
1--2--3--4--5--6--7--...

Well 1-- all comes from the first film frame. It's a nice clear image from a single film frame. 2-- is also a nice clean frame. But 3-- is partly from frame 1 and party from frame 2 it looks like a double exposure of film frames 1 and 2. This is what happens with field blended frame rate conversions. Some of the fields contain only data from one film frame but some contain double exposures of two film frames. These are a little harder to restore and you will usually still have a few double exposure frames after restoration.

You can use some thing like this AVS script:

Load_Stdcall_plugin("C:\Program Files\AviSynth 2.5\plugins\yadif.dll")
Import("C:\Program Files\AviSynth 2.5\plugins\SRestore.avs")

MPEG2Source("1st-amb.d2v")
Yadif(Mode=1,Order=1)
SRestore(FRate=23.976)
LanczosResize(720,480)

Here's the first clip after that script:
1st.avi

Regarding the shimmering on the door, I assume you mean the way the thin lines of the panels buzz as the door is moving. Those are aliasing artifacts from poor resizing -- part of the original NTSC to PAL conversion. If you play around with different deinterlacing (ie, Yadif vs LeakKernelDeint, etc) and resizing algorithms you might be able to make some improvement. Or they may get worse.

[StrobeLightEpileptic]

Interesting discussion. I'll have to dwell on this. But I did try your script.

Yes, it does improve things quite a bit. I loaded it into Vdub to see, but when I loaded the script in tmpgenc I began getting an error message on the tmpgenc video window .

The error reads verbatim: "I don't know what "d32" means. ([ScriptClip], line 19)"

Well, that took me back to earlier this year when we had a discussion about blended frames. A script was suggested using SRestore and I kept getting the same error message in TMPGenc. I don't know what I did back then to correct the problem. Tonight I downloaded the most recent version of SRestore to see if it made a difference, but I keep getting that error.

But the script works fine in VDub.

[manono]

And don't forget to greyscale it. Joseph Cotten looks a little greenish.

As for the script, see if it loads into HCEnc. It's better and faster anyway. Or make a lossless Lagarith AVI from the script using VDub and then load that into TMPGEnc.

[StrobeLightEpileptic]

And don't forget to greyscale it. Joseph Cotten looks a little greenish.

As for the script, see if it loads into HCEnc. It's better and faster anyway. Or make a lossless Lagarith AVI from the script using VDub and then load that into TMPGEnc.

Yeah, that does look a little off color. But, I got to trying to remember what I did earlier this year, and remembered the VFAPI priorities in tmpgenc. There is one of those plugins that is used for d2v files and when I gave it a higher priority the error message went away.

So, now I'm trying this encode with tmpgenc, but I'll remember your suggestion.

[StrobeLightEpileptic]

I am just tonight able to get back a response to this thread. Jagobo, I have to say that the result of this encoding is great. There is no jerky movement at all and you have to look pretty hard to find any of that shimmering, (and when I think I see it I don't know if it's just my imagination). This is an instance where the IVTC gave a result that is better than the source.

So, a person does need to be skilled at recognizing what has been done to a film before they can apply the correction. So, if I get what you have described, "out of phase" video happens mainly with a capture card. I guess a person just needs to see the fields "stepped" to recognize it?

And blended frames can be recognized by fields that look double exposed? And this is caused by the different frame rates that are used converting the film to video.

progressive is the easiest to understand. And I don't have much of a problem understanding NTSC interlaced.

But, I am having a hard time understanding PAL interlaced in light of this discussion. When technicians interlace video for pal, do they simply add two fields/sec? So a person has to write down the fields:

A B C D E E F G H I J K L M N O P Q R S T U V W X
A B C D E F F G H I J K L M N O P Q R S T U V W X
----------------
Or just what is the method for properly interlaced PAL video?

Regarding the script: This isn't the first time you Manono have suggested SRestore and Yadif. I understand the 2nd parameter for TFF or BFF, but I have not been able to understand what "spatial interlacing" is. There is very little on the web about it, (that I could find anyway).

Anyway. I hope to apply of this on future films I want to work with.

Thanks.

[manono]

This is an instance where the IVTC gave a result that is better than the source.

You didn't IVTC it; you unblended it. It was never telecined and thus it can't be InVerse TeleCined.

So, a person does need to be skilled at recognizing what has been done to a film before they can apply the correction.

Yes, you have to know what you have before you can know how to treat it.

So, if I get what you have described, "out of phase" video happens mainly with a capture card.

It's seen in PAL DVDs as well.

I guess a person just needs to see the fields "stepped" to recognize it?

If the frames appear interlaced but after separating the fields the fields are in duplicate pairs, it's got shifted fields (out of phase).

There's something similar, though, where you have 12 or thirteen frames interlaced followed by 12 or 13 'clean' frames. Half are phase-shifted and half aren't. In that case it was converted from film to PAL by adding a frame every second in the form of a field every half-second.

And blended frames can be recognized by fields that look double exposed?

Yes, after separating the fields or bobbing the video many of the fields appear as double images, ghosted, blended.

And I don't have much of a problem understanding NTSC interlaced.

Me, I find that quite complex, since there are several kinds including true interlaced video, hard telecine, and several kinds of field-blending, all leading to interlaced frames and interlaced encoding.

When technicians interlace video for pal, do they simply add two fields/sec?

If you're talking about phase-shifted PAL, then it's more like:

A B C D E F G
b c d e f g h

If they're converting from Film to PAL, it might be something like you diagrammed, but the added fields are spaced a half second apart, and not exactly as you diagrammed it. The idea is to make it play less jerky.

[jagabo]

"out of phase" video... a person just needs to see the fields "stepped" to recognize it?

A quick way to recognize out-of-phase PAL is to Bob() the video. If frames look interlaced before the Bob() but every pair of fields looks alike (except for the usual Bob artifacting) then it is out-of-phase PAL. You can use VirtualDub's Bob Doubler filter or AviSynth's Bob() or any of the smart Bobs, Yadif(mode=1), for example.

Another quick test is to:

SeparateFields() #separate frames into fields
Trim(1,0) #throw away the first field
Weave() #recombine the fields into frames

If all the frames look progressive after that you know it's out-of-phase PAL.

just what is the method for properly interlaced PAL video?

There are three main ways to convert 24 fps film to 50 fields per second video:

1) Speed up the film and audio. Ie, just "project" the film at 25 fps. Then you display each frame as two fields. The running time of the movie is about 4 percent shorter and the audio pitch is about 4 percent higher (unless a pitch shifter is used).

2) Duplicate one frame out of every 24, thereby creating 25 frames out of 24. Then you display each frame as two fields. The running time of the video and the audio are unchanged. But there is one jerk every second because of that duplicate frame.

3) Duplicate two fields every second. This creates 25 frames from the original 24:

A B C D E F G H I J K L L M N O P Q R S T U V W X 
a b c d e f g h i j k l m n o p q r s t u v w x x

You can see that Frame L has become three fields. Later frame X becomes three fields. Two extra fields out of every 24 frames creates 25 frames (50 fields). The running time of the audio and video is unchanged. You will see two little jerks every second. If you look at frames you'll see half the frames look progressive and half look interlaced, switching between the two every half second.