TIVTC with "external" telecine pattern

[mp3dom]

Hi to all!
I have a question regarding TIVTC and telecine in general...
I have an NTSC video coming from an NTSC DVD that is progressive (DGIndex says 98% Film) and I can easily IVTC with TIVTC. I assume the pulldown pattern is written in the MPEG (trf, rff flags) so TIVTC and also DGIndex known what are the interlaced frames and "good frames" without problems. Now I have the same source/video but this time coming from a Digibeta. Unfortunatly with this source I lose the mpeg flags (infact DGIndex this time says it's "Interlaced" at 29.976 fps). The question is... assuming that both video starts/ends at the same frames is it possible to pass the "ntsc dvd" pattern to TIVTC and IVTC the "Digibeta" source? What I want to avoid is to TIVTC the digibeta source and obtain some interlaced stuff/bad ivtc that I can easily avoid using the dvd source that is progressive at 23.976fps...
Thanks!

[Guest]

If it's the same video/source as you say, why not just use the DVD source?

And if the DVD is 98% film, don't use external IVTC, use Force Film instead.

[mp3dom]

Thanks for the reply neuron!
I would like to use the digibeta source simply because it's better in quality over the dvd...(I need to convert to PAL and this time I would like to avoid the frameblending since the source is true film) but the "video" (image) is the same... so the idea would be to obtain the best from both... the "pattern" from the dvd and the quality from digibeta.
If I "Force Film" can I risk to have some combed frames? (the remaining 2%)
Thanks!

[Guest]

The video will have to be exactly the same all the way through for this to work. How likely is that?

Ask tritical whether TIVTC supports this. I think it may support this.

The 2% is likely just the movie lead-in or credits. You can catch those with postprocessing with a relatively high threshold.

[mp3dom]

Thanks! I'll ask to Trictical. Thanks again!

[jmac698]

I've had this problem as well, two sources, one progressive. Same reason as well, the interlaced source was much higher quality. And yes, they originated from exactly the same source. If there was any editing, it was likely only at the scene level.
I looked into it, and there's an external ivtc file accepted, however it's limited. (Through the overrides file).
See
http://forum.doom9.org/showthread.php?t=82264

[mp3dom]

Thanks! I've searched a bit in the TFM documentation and I've found the "output" parameters that output a file with the IVTC frame decision. This file could be specified with "input" parameter on the interlaced source to obtain the same result (if both interlaced and progressive+pulldown were the same). To be sure there are no differences I'm IVTC both sources with output parameter and then compare the 2 files to see where frame decision change.

[Mug Funky]

is it not possible to capture the didge in another format? mpeg-2 isn't ideal considering the source is 90mbps 10 bit 4:2:2.

if you can spare the space (and have the gear), you should be able to cap into a 2Vuy mov file with all 4 audio channels at 24 bits. this can be loaded into avisynth (though the audio will have to be treated separately) and TIVTC'd at possibly greater precision than the encode you have on DVD.

[mp3dom]

Sorry Mug Funky but I don't have understand very well your post.
I can capture also in Avid Meridien Lossless but I'm unable to open it in AviSynth in a good way. I've tried the QTInput but I can work on it only if I output to RGB colorspace (and I would like to avoid it...). Anyway, capturing in uncompressed mov or mpeg at 50mbps (always 10bit 4:2:2 through SDI interface) doesn't make so many differences (excluding the mpeg compression) because both captures are "hard pulldowned". The only source with soft 3:2 pulldown is the dvd source... here the request to have the pattern from the dvd source applied to the hard pulldown source.

[Mug Funky]

hard pulldown @ 50mbps beats soft pulldown @ max 9.8mbps by a long way...

i didn't realise your mpeg-2 captures could go that high. the encoders i've got plugged into the decks here only go up to 9.6mbps (they're for DVD), and actually allow pulldown detection (soft pulldown).

there's probably not much to be gained from using the pulldown flags in the original source, as you can probably get better matches with tfm.

i believe tfm also allows the loading of a d2v file as reference, so provided you sync your capture up to the DVD source, you'll be able to do it that way.

however, there's an advantage to capturing in blackmagic 8 bit (2Vuy - the capital "V" is important, too) and loading it with:
QTinput("buh.mov",mode=1,raw="uyvy",audio=true). this will return the raw data captured, and will give you a big speed advantage compared to using quicktime itself to decode (right now QT is used as a sort of splitter, but it serves the raw UYVY to avisynth).

[Trixter]

i didn't realise your mpeg-2 captures could go that high. the encoders i've got plugged into the decks here only go up to 9.6mbps (they're for DVD), and actually allow pulldown detection (soft pulldown).


Sounds like a BlackMagic Design card. My last DVD project was done entirely with a $500 DeckLink SP card -- supports 4:2:2 capture at 10-bit...

[Mug Funky]

they don't do mpeg-2 do they? i'm reasonably familiar with the decklink HD cards, and i've never seen an option to use anything but what quicktime says i can use. of course for avisynth blackmagic 8 bit is enough.

i wish someone would port huffy to quicktime - massive space saver, especially with anime which is mostly flat. and 10-bit support would be sweet, too :)

[Trixter]

they don't do mpeg-2 do they? i'm reasonably familiar with the decklink HD cards, and i've never seen an option to use anything but what quicktime says i can use. of course for avisynth blackmagic 8 bit is enough.


Yep, I used 8-bit, although there was noticable (as in, you could see it in the images) loss if avisynth converted to YUV before MPEG-2 encoding. To preserve quality, I kept things RGB and fed RGB into CCE with a 10-bit DCT. At 10-bit, the RGB->YUV conversion was nearly lossless (no coarse gradients, etc.)

The DeckLink software package doesn't support MPEG-2, but it does support realtime JPEG now, so you can have realtime 4:2:2 capture but only taking up between 3-10MB/s depending on your quality settings. (Uncompressed 8-bit 4:2:2 was 25MB/s)


i wish someone would port huffy to quicktime - massive space saver, especially with anime which is mostly flat. and 10-bit support would be sweet, too :)

There are already lossless, HDR, etc. codecs for quicktime -- but they cost money, of course :-)

[Mug Funky]

Yep, I used 8-bit, although there was noticable (as in, you could see it in the images) loss if avisynth converted to YUV before MPEG-2 encoding.

i was talking about capping in 8-bit yuv. no loss feeding that into cinemacraft. of course, if quicktime is serving it up in RGB then you've already lost a lot if the source is 8-bit yuv. the trick is using QTsource's raw mode so that you can guarantee no colour conversion is happening (QT also b0rks YUV to RGB conversions because it appears to use it's own set of coefficients. i'd rather let avs deal with that stuff).

you could also cap in RGB natively, but there's no point considering SDI is 4:2:2 yuv, and even with 10-bits the upsampling to RGB is lossy.

...unless you're lucky enough to have dual link stuff.

[Trixter]

i was talking about capping in 8-bit yuv. no loss feeding that into cinemacraft.

No, but it can be a pain if you have to edit (which I did). My choices were:


Capture YUV and hope I keep things to 16-235; only 220 luminance levels to work with
Capture RGB, let the editing software hard clip to 0-255; 255 luminance levels to work with


I had more range to work with RGB (I had to do some color correction, noise filtering, matting, other stuff), so that's how I captured and edited, and with a 10-bit DCT, no loss during the 0-255 -> 16-235 conversion.

For anyone who wants to check my work, I urge you to check out www.mindcandydvd.com ;)