Seeking AviSynth script impr0vement suggestions

[jman98]

I've got a Hauppauge Colossus card that I recently purchased and I've been doing some laserdisc captures with it. It records in H.264 for video and the audio is AAC. I can convert the audio without any issues. I tried the below script and it did not give good results. Essentially motion looked jerky. I hardly claim to be an AviSynth expert. I've done research on my own and there's a real lack of examples out there so the below represents my best guesses. I'm certainly open to any suggestions for improvement. My desired goal is convert interlaced 59.94 fps input to 24 fps progressive output. Except for the unnatural jerky look to the motion, the quality of what got produced by the script was fine for me. Here's the script:

LoadPlugin("C:\megui\tools\dgavcindex\DGAVCDecode.dll")
AVCSource("F:\WD-WWOC\Dis-complete_track1_und.dga")
LoadPlugin("C:\megui\tools\avisynth_plugin\LeakKer nelDeint.dll")
LeakKernelDeint(order=1,sharp=true)
LoadPlugin("C:\megui\tools\avisynth_plugin\Decomb. dll")
AssumeTFF()
Telecide(guide=1)
Decimate(5)


Suggestions for improvement and corrections of mistaken assumptions on my part would be welcome.

Finally, I am just making a wild guess on TFF with my video. Since it's H.264 I do not know how to tell which field the interlaced captures use first. Any suggestions? I've got tools for MPEG-2 that can tell me this, but I don't have any tools for H.264 that can tell me. I have been unable to determine from looking at various forums which field the card records first as the people that own it don't seem to be interested in that subject.

[Selur]

personally I use QTGMC for deinterlacing, TIVTC for telecine stuff and FFMpegSource2 to load source material.

Suggestions for improvement and corrections of mistaken assumptions on my part would be welcome.

you asked for it

Essentially motion looked jerky.

1. you go from 59.94*2 ~ 120 fields with motion information down to 24 fields with motion information,..
-> I don't think it's possible to do this without loosing motion information and thus cause some sort of motion looking jerky of blended frames
2. if your display does not support 24p natively you also will end up with some sort of jerky motion,..
-> I would recommend to do some basic search over at doom9 (there are a bunch of threads that deal with 60i to 24p conversions and the possible solutions,..

Finally, I am just making a wild guess on TFF with my video.

-> read the Decomb documentation (you use the filter, so I guess you got the doc), it describes how to figure out if you content is tff or bff in the tutorial

regarding your script:
I'm unsure why you do what you do.
First you deinterlace you material to 60p (using LeakKernelDeint), when you to a IVTC with a NTSC 3:2 pulldown pattern guidance on material that at least after the deinterlacing could not be called telecine,..

-> all in all I would recommend to look at the motion blur&co solutions over at doom9
(a simple search on google for:
60i to 24p avisynth
should also show a bunch of probably better working scripts)

Cu Selur

Cu Selur

[jman98]

regarding your script:
I'm unsure why you do what you do.

Ha ha - I admitted my understanding was NOT perfect! That's why.

THank you very much for the reply. I will look into all of those things.

[poisondeathray]

Is the laserdisc content 60i ? It's encoded as 60i, but is the content actually 60i ?

I don't understand why you are doing this either? If it's film content, then you would IVTC and it will be 23.976. If it's interlaced content, then you could deinterlace it to 59.94p or 29.97p .



You can determine the field order using this miniguide (instead of MPEG2Source , use something else for source filter)
http://neuron2.net/faq.html#analysis

Note DGAVCIndex is depreciated, no longer supported by the author, and has problems with PAFF streams such as you are recording with Hauppauge card

[jman98]

Is the laserdisc content 60i ? It's encoded as 60i, but is the content actually 60i ?

I don't understand why you are doing this either? If it's film content, then you would IVTC and it will be 23.976. If it's interlaced content, then you could deinterlace it to 59.94p or 29.97p .



You can determine the field order using this miniguide (instead of MPEG2Source , use something else for source filter)
http://neuron2.net/faq.html#analysis

Note DGAVCIndex is depreciated, no longer supported by the author, and has problems with PAFF streams such as you are recording with Hauppauge card

60i is what the damn card gives me - as best I can tell. I really wish the card would record in MPEG-2 because that's a ton easier for me to deal with than H.264. Since the source is laserdisc, I assume the source is actually 29.97i.

I appreciate this and any other feedback. In the end I may just go with the 60i as is but I'm still interested in any further points anyone else wants to make. Thanks for the info.

[poisondeathray]

60i is what the damn card gives me - as best I can tell. I really wish the card would record in MPEG-2 because that's a ton easier for me to deal with than H.264. Since the source is laserdisc, I assume the source is actually 29.97i.

I appreciate this and any other feedback. In the end I may just go with the 60i as is but I'm still interested in any further points anyone else wants to make. Thanks for the info.

Sorry, I don't know a damn thing about laserdisc. Are you saying theatrical movies on laserdisc are actually interlaced CONTENT ?

In DVD's and MPEG2, they are 23.976 film content. They are telecined (either hard or soft) to 60i. The DVD signal is 60i, but the CONTENT is 23.976 for theatrical movies. For sports DVD's, the content can be 60i (true interlace)

A hardware encoder like Colossus will only capture the signal as 60i, but if it's actually 23.976 content, you have to remove pulldown (inverse telecine). If you captured 1080i60 TV show from HDTV capture, it's analgous to that - the signal is 60i, but the content is 23.976 progressive . But a sports HDTV cap would be 60i signal, 60i content (ie. interlaced content)

If you post a sample I'm sure we can sort it out

[jman98]

Here's a 10 second sample. I'm just assuming it's film since it was filmed in the 1950s. Maybe it's not film as it was for TV (although the original source is still film). Would it make more sense to just leave it interlaced (I'm OK with that) or should I convert it to 29.97 fps progressive?

[poisondeathray]

Here's a 10 second sample. I'm just assuming it's film since it was filmed in the 1950s. Maybe it's not film as it was for TV (although the original source is still film). Would it make more sense to just leave it interlaced (I'm OK with that) or should I convert it to 29.97 fps progressive?

It's film rate, you just need to inverse telecine

e.g.

FFVideoSource("Dis-sample.mkv")
AssumeTFF
TFM()
TDecimate()

How did I determine this? Just go to that link above, and run through the exercises. You will see 3:2 pattern

ffms2 is external plugin (doesn't come with standard avisynth install)
http://code.google.com/p/ffmpegsource/

[manono]

60i is what the damn card gives me - as best I can tell. I really wish the card would record in MPEG-2 because that's a ton easier for me to deal with than H.264. Since the source is laserdisc, I assume the source is actually 29.97i.

60i and 29.97i are exactly the same thing, one referring to fields, the other to frames.

Had to edit as I was getting one thing from playing the video through MPC-HC, and another playing a script using FFVideoSource.

Yes, as pdr says, IVTC it.

...or should I convert it to 29.97 fps progressive?

IVTCing it will convert it to progressive 23.976fps, film rate. Don't ever deinterlace a film source.

[poisondeathray]

I think Colossus records in transport stream, but he just re-wrapped it into mkv

[jagabo]

Sorry, I don't know a damn thing about laserdisc. Are you saying theatrical movies on laserdisc are actually interlaced CONTENT ?

Laserdisc is simply a composite video signal recorded on to disc. Output is always 29.97i. The Colossus is recording 29.97i. 59.94i is just the new name for 29.97i, foisted upon us by marketing droids.

[Selur]

59.94i is just the new name for 29.97i,

Not really a new name,.. more like a conflict between how people tend to define streams.
Some always speak of fields others of frames, normally it kind of depends on the context and the lingo that is custom in it, problem is the whole thing get's mixed up.
I agree with poisondeathray using IVTC should be fine, if one uses:

Decombs:
Telecide(guide=1)
Decimate(5)
or TIVTCs:
TFM()
TDecimate()
is probably more a matter of preference. (both should do the job)

[poisondeathray]

59.94i is just the new name for 29.97i, foisted upon us by marketing droids.

Yes, I caved in.

I used to call it 29.97i too, but virtually all recent programs, cameras, manuals in the last few years call it 59.94i now (60i to be lazy)

[Selur]

ATSC uses fields, see: http://www.hdtvprimer.com/ISSUES/what_is_ATSC.html, a lot of standard always used another naming sceme like 1080i/25 (1080i with 25fps; so basically 1080@50i ) and then it got all mixed up and you get 1080/50i (http://www.avchd-info.org/format/index.html), some use frames some use fields as basis and everybody is confused.
(marketing prefers 50i over 25i since bigger numbers must mean better sales )

[jman98]

Sorry for the delay. Things were a little crazy for me yesterday and between some personal things and some after hours work I had to do, I didn't have any time to look into this again until today.

I want to thank everyone who posted and corrected my misunderstandings. I've got one of Selur's suggestions working now and it looks like it can do what I want, but all the posts were helpful to me so thank you again to all who responded.