AviSynth with VirtualDubMod - Inverse Telecine advice needed

[spicediver10191]

Hello,

Could I please ask the advice of the AviSynth experts here?

I'm encoding a large uncompressed AVI (40GB, progressive, 29.97fps) to Xvid using VirtualDubMod.

I have an AviSynth script that already produces superb results with AviSynth version 2.5.8:

#ASYNTHER Default (AVISource)
AVISource("E:\XYZ project\Uncompressed AVI exports\Sequence 01 (lagarith RGB).avi")
ConvertToYV12()
RemoveGrain()
Tweak(sat=1.15)
Lanczos4Resize(720,304,0,60,0,-56)
LimitedSharpen()

However, I also want to encode a 23.976fps Xvid version as well, so I need to do an Inverse Telecine, 3:2 pulldown thingy. It seems using the Avisynth Decomb filter is the best way to do this.

I have 3 questions about using Decomb:

1. Because my source AVI is already progressive, do I even need to use the Decomb commands AssumeTFF() and Telecide()?

2. For framrate conversion down to 23.976, is simply typing "Decimate()" in the script all I need, or should I add some perimeters to that command?

3. At exactly what stage/line/order in my script above is the best place to insert this task/tasks?

If you could post an example Avisynth script for me that'd be grand.

Thanks.

[manono]

If the source is true 29.97fps (every frame different and unique), then decimating it to 23.976fps will make it play jerky/jumpy. Forget the idea.

Or is there already one duplicate frame in every 5-frame sequence?

[spicediver10191]

If the source is true 29.97fps (every frame different and unique), then decimating it to 23.976fps will make it play jerky/jumpy. Forget the idea.

Or is there already one duplicate frame in every 5-frame sequence?

No I don't think so. I've just examined a panning shot frame by frame in VirtualDubMod and I can't see any repeated movement. I don't know if VDM has a diagnostic for detecting such frames or not. I'm just going by my eyes.

The footage is a compile of a number of progressive NTSC DVD sources that were frameserved and edited in Premiere CS3 as a 29.97fps project, and then exported at the same fps.

The reason I'm doing the conversion is only because someone I gave the file to said that 23.976 fps will make the file more friendly for distribution, specifially for playback on some computers. Is he even right, I wonder?

[manono]

Is he even right, I wonder?

No, and if he thinks decimating true 29.97fps video to 23.976fps is a good idea, he's not only wrong but ignorant.

[AlanHK]

The reason I'm doing the conversion is only because someone I gave the file to said that 23.976 fps will make the file more friendly for distribution, specifially for playback on some computers. Is he even right, I wonder?

If the video was made with pulldown, (i.e., repeated frames) as most NTSC TV shows and movies are, yes, it would give better quality. But if not, no.

Also, your script:

...
Lanczos4Resize(720,304,0,60,0,-56)
LimitedSharpen()

You can use LimitedSharpenFaster (aka LSFmod) and do the resize as part of that:

Crop(0, 60, 0, -56)
LSFmod(strength=100, dest_x=720, dest_y=304, defaults="fast")

Adjust strength to taste.

see http://avisynth.org/mediawiki/LimitedSharpen, the AVSI is linked from there.

[spicediver10191]

Thankyou both.