Advice needed, converting laserdisc (29.97fps) to progressive 24fps

[Jeff D]

I'm going to start with what I'm looking to do, because I'm sure some will comment on other setups.

I'm looking to make a 24fps file for dvd from laserdiscs. The display is a HDTV progressive 16:9 widescreen display. (I don't care about 4:3, iterlaced, etc)

I've done searches, tests and read til my eyes are sore, the closest solution around here is:
http://forum.doom9.org/showthread.php?s=&threadid=17792

My setup:
P4 2.4Ghz, 1GB Ram, Lots of fast hard disk space.
ATI AIW 8500 128MB cap card, and Santa Cruz sound card for audio cap.
Promise UltraATA100 controller and more capture disk space.
Capture with vdub in HuffYUV (best), (best) settings for the two drop down configuration options. The defaults I believe.

First step:
Good capture, I've used vdubVCR mod to get cap each of the 3 discs with no frame drops. (I'm needing to be sure this app doesn't add frames, for that will screw up the IVTC step later). Caps were 29.9706 fps (closest I could get to 29.97)

I've been trying to get a good IVTC done with the captures and it's not working.

I've been using Don's Decomb filter with avisynth 2.5 and it's ok. But, it's not correct. The Telecine filter is followed by a decimate.

Telecine modes tried:
Default, blend=TRUE (default)
Default, blend=FALSE
guide=1, blend=TRUE (default)
guide=1, blend=FALSE

The first SW disc get to the point where Leia's ship is hit and we cut inside to 3po and r2d2. The camera shake causes blur and movement that confuses the telecine filter. The results look like two fields were mismatched and combined. Two images of 3po side by side.

This clip doesn't seem to be an easy one for the filters to work with.

Other problems include there are scenes where it appears the capture was done is 16bit color depth. I can post images if that would help. The desktop was set to 16bit, but I thought the cap settings were independant, althought I can't see where to set that. Another question, is there a way to force the cap card to do a top or bottom field first on the cap?

I have about 20 test clips and notes on different settings I've tried with vdub filters and avisynth filters. I think avisynth is the way to go for customization.

My future plans include to resize the image to an anamporphic size for widescreen playback. But, before that I need to get the telecine and decimate working well.

[Jeff D]

I forgot to mention I've also tried the code Atlanis posted in the linked thread.

I just did the doubleweave and the result is still interlaced looking.

I don't see how, but could the source be really screwed up? Anyone know how to know if vdubVCR does something funky with extra frames?

[Jeff D]

I'll keep replying to my posts, maybe someone else will join in.

I believe the source, SW definitive set, is pretty crappy. There are frames during cuts, the explosion on leia's ship to inside with 3po and r2, where the interior shots have a slight coloration of the explosion, that disappears a couple of frames later.

I've got more questions for the experts, you're getting behind... =)

I've been playing around with the fields and used the following script from the avisynth website:

function CheckTopFirst(clip v1) {
v2 = v1.subtitle(v1.GetParity()?"TF":"BF")
v3 = v2.GetParity()?v2.ComplementParity():v2
t0 = v3.ComplementParity()
t1 = t0.separatefields()
t2 = compare(t1.trim(1,0),t1)
b1 = v3.separatefields()
b2 = compare(b1.trim(1,0),b1)
return stackvertical(t2,b2)
}

v=OpenDMLSource("G:\CAPTURE005.avi")

CheckTopFirst(v)


I have a question about this? When run the image on the top and bottom seem to flip, for example every other line of the BF appears int the top half, starting with the top of the BF text. the bottom image shows every other line of the BF starting at the second line of the text. Advance forward one frame and I see the exact opposite, the top image has every other line of BF starting with the second line and the bottom has every other line of BF starting on the first line.

Does it make sense for the script to be doing that? Shouldn't the text, although every other line, be solid with the first row and every other row showing up always in one half of the shown image with the other lines showing in the other have of the shown image?

Examples of alternating text (files are big, no photo edit sw on the capture PC):
Frame0 (http://home.attbi.com/~jjd59/frame0.bmp)
Frame1 (http://home.attbi.com/~jjd59/frame1.bmp)

[Atlantis]

Hi!

2 possible problems causing this.

First about capture. It could be VirtualDub. I have tried the latest VD for capture and simply I was never successful with VD. First it gives wrong frame rate. I have no idea why this wonderful program is unable to capture at a constant 29.97 frame rate. It’s always something else. I also get a lot of dropped frames. So I don’t use VD. I use AVI_IO for capture. It gives an accurate frame rate count and no dropping!

2) the old script was:

OpenDMLSource("star_wars.avi")

DoubleWeave()

#Pulldown(0,2)
#Pulldown(1,3)
Pulldown(2,4)
#Pulldown(0,3)
#Pulldown(1,4)

which works perfectly for DV capture. But I have noticed that it might not work for other analog capture cards and the solution is this:

SegmentedAVISource(“------“) if captured with AVI_IO

AssumeTTF()
#AssumeBFF()

DoubleWeave()

#Pulldown(0,2)
#Pulldown(1,3)
Pulldown(2,4)
#Pulldown(0,3)
#Pulldown(1,4)

You should try AssumeTTF() or AssumeBFF() but not both at the same time. So in short you will have 10 different combinations that you have to try. AssumeTTF() or AssumeBFF() is related to your capture card. Once you find the correct setting, you will keep it forever for that card. But the other 5 options should always be tried for each side of your LD captures.

Also don’t forget to try in VD both the first and end of your each LD side to make sure it’s progressive. Also don’t forget you should get 23.976 as the frame rate and not 24! Forget the 24! You want 23.976 for NTSC on DVD.

[Jeff D]

Thanks Atlantis, I tried a sample clip with the AssumeTFF and the last pulldown settings worked.

I'll try the BFF just to be sure tonight.

I've seen a AVI_IO mentioned a lot around here. And it's been recommend by others on the AVSForum. But, I'm concerned about the addition of frames during capture and what that could do to the IVTC stage.

And yeah 23.97, I round to 24 for simplicity. Guess that might be a bad idea, details...

BTW Everyone knows film is shot 24fps, but did you film is actually projected at 48fps with each film cell being shown twice? =)

[Jeff D]

Atlantis, I tried the following combos on two of my 3 big caps, and couldn't get one of them to work right. (I think I may have forgot the DoubleWeave call on the sample clip I mentioned above)

v=OpenDMLSource("G:\CAPTURE003.00.avi")

v1 = DoubleWeave(v)
v2=AssumetFF(v1)

#v3 = Pulldown(v2,0,2)
#return subtitle (v3, "0,2")
#v3 = Pulldown(v2, 1,3)
#return subtitle (v3, "1,3")
#v3 = Pulldown(v2, 2,4)
#return subtitle (v3, "2,4")
#v3 = Pulldown(v2, 0,3)
#return subtitle (v3, "0,3")
v3 = Pulldown(v2,1,4)
return subtitle (v3, "1,4")


Is there anything that I'm doing wrong? I was planning on finding the right unique combo for each of the files, but I can't even get one combo right. Ten attemps at each, and nothing.

I was able to get some other sample caps to work correctly, but that was with a different version of vdub, not the vdub with VCRmod. The big caps were done captured with the VCRMod version.

[Atlantis]

Look at my script. AssumeTTF() comes before DoubleWeave().

[lancer]

personally I would suggest you use the AVS decomb filter script instead of pulldown and weaving the fields together. I got much better results.

[wotef]

you may have to resort to telecide manual overrides - it's tedious but it will get the job done, i believe the only field matching patterns needed are N and P

[Jeff D]

I did some more testing and with my caps I found Atlantis' method way better than the avs telecine and decimate from Don's Decomb filter.

With Don's filter I tried a few settings.
Telecide().Decimate(5) - results frames that would blur in and out almost on a frame basis.

Telecide(blend=false).Decimate(5) - blend seems to keep the image from bluring in an out, but the image quality isn't right, edges are very strange looking, stair stepping on diagonal lines and such.

Telecide(guide=1).Decimate(5) - same images bluring in an out.

Telecide(guide=1, blend=false).Decimate(5) - (same as blend=false above)

I'm open to other suggestions on settings for T&D calls, I'm sill learning... =)

I found the Assume?FF, doubleweave and pulldown to produce a very nice image. Very nice for a while then maybe the cadence of the frames changed or a dropped frame did something that caused the sequence to comb. There were dropped frames, and I'm suspect of that, but I'll try to recap and see how that looks.

I can post images if anyone is interested in what Don's filters were doing.