How to read a d2v file

[real.finder]

CD2VPropFix

dgindex already has fix option that on by default

https://i.postimg.cc/1z54Kw8Y/Screenshot-2022-04-19-222331.png (https://postimages.org/)

[Ceppo]

The name mislead you, run it and use PropShow() :)

For TFF, I mean that we don't know if it is TFF or BFF if Field = 0. And on the other forum the author said that you need both flags for the job. And more we have the better :)

[real.finder]

The name mislead you, run it and use PropShow() :)

I see, but why almost all frames are PureInterlaced?

For TFF, I mean that we don't know if it is TFF or BFF if Field = 0. And on the other forum the author said that you need both flags for the job. And more we have the better :)

maybe the new RFF property can be set to either 0 which mean no RFF or 1 which mean RFF with TFF or 2 which mean RFF with no TFF, edit: or 1 which mean RFF with no TFF or 2 which mean RFF with TFF

[Ceppo]

I see, but why almost all frames are PureInterlaced?
If all frames of the cycle are flagged with fieldbased > 0, the cycle is pure interlaced. Still you need the iiiii pattern to get a 60i. Otherwise is uuuuu.
If you are using my pokemon source that is, if not please share the source.


maybe the new RFF property can be set to either 0 which mean no RFF or 1 which mean RFF with TFF or 2 which mean RFF with no TFF, edit: or 1 which mean RFF with no TFF or 2 which mean RFF with TFF

Either we do 3210 flag as it is meant to be read or we get them both separate. I prefer the latter.

[Ceppo]

//

Fix'd, I replaced get line with ifile >> line; go figures how getline works with that much amount of data :)

[Ceppo]

@real.finder
any luck with the new d2vsource? I don't know if it is my source but I get strange flags.

[real.finder]

@real.finder
any luck with the new d2vsource? I don't know if it is my source but I get strange flags.

how strange?

[Ceppo]

Like two consecutive "3" and such which don't make any sense :(

[real.finder]

Like two consecutive "3" and such which don't make any sense :(

maybe ask dg for help? http://rationalqm.us/board/viewtopic.php?f=7&t=1157&p=15628#p15628 I think he may be interested also in adding these frame properties to dgsource

[Ceppo]

He already got the skill to do it if he wanted. He doesn't need a nub help (me).
Also, I think the problem is if we honor the pulldown the flags get confused since in the source there are 4 frames, if there are 2 RFF flags 2 fields get repeated and you get the fifth frame and you get your 29.970 telecined source.

[real.finder]

I don't think he cares. He already got the skill to do it if he wanted. He doesn't need a nub help (me).


I didn't mean you help him but he help you :)

I mention dgsource since no one but him can update it since it closed-source, also as incentivize for him to help you more and make it common (standard) so other plugins will take the same path

Also, I think the problem is if we honor the pulldown the flags get confused since in the source there are 4 frames, if there are 2 RFF flags 2 fields get repeated and you get the fifth frame and you get your 29.970 telecined source.

this is not a problem rather it is the normal case, maybe you can take some ideas from tfm?

[real.finder]

ok I did try play with it, according to my understanding of http://rationalqm.us/board/viewtopic.php?p=15627&sid=efe069d7b13d400d15d9e4d4367f6223#p15627

this should do it
D = DoubleWeave()
M = CTelecine()
Q = M.CPostProcessing()
ScriptClip("""
ATOP = PropGetInt("_EncodedFrameTop")
ABOT = PropGetInt("_EncodedFrameBottom")
FIELD = PropGetInt("_FieldBased")
rff = PropGetInt("_RepeatedFirstFrame", offset=-1)
tff = PropGetInt("_TopFieldFirst", offset=-1)
FIELD != 0 ? Q.Subtitle("Not Soft",400) : \
ATOP != ABOT ? D.DuplicateFrame(0).SelectEven().Subtitle("Soft Match",400) : last.Subtitle("No Match",400)
FIELD == 0 && IsCombedTIVTC() ? M.Subtitle("Bad Soft",400) : last
rff == 1 && tff == 1 && FIELD == 0 ? propSet("_ToDecimate", 1) : propSet("_ToDecimate", 0)
""")

but d2vsource _RepeatedFirstFrame (typo (https://github.com/Asd-g/MPEG2DecPlus/issues/9#issuecomment-1105880684)) and _TopFieldFirst seems need more work to do it :) since it kinda random I think...

btw, DG seems care about this, just try and ask him in his forum, he told me he will not post in doom9 anymore

[Ceppo]

@real.finder
The true course of action is to use ignore pulldown and detect VFR and do field matching/deinterlacing for post telecine editing.

The last D2V source, with ff=true, seems to do the job, I will make a plugin to read those and see if I can detect correctly VFR with the available props. Then I will post in his forum to see if he wants to update the DG family (in the case I'm successful).

[gonca]

Then I will post in his forum to see if he wants to update the DG family (in the case I'm successful).

I think he is already aware, if you check

[Ceppo]

Thanks, I didn't check. He will probably do a better job than me.

[gonca]

You should keep working on your plug in
I am sure that each will have its advantages
And he is still asking for feedback

[Ceppo]

I run into a problem, I'm unable to get an accurate timestamp, and I'm always off of a few seconds. The problem is how to handle the following cases:

3 28
0 29
1 30
2 31
3 32
0 33
1 34
2 35
3 36
0 37
1 38
2 39
3 40
0 41
1 42
2 43
3 44
0 45
1 46
2 47
3 48
0 49
1 50
2 51
2 52
2 53
2 54
3 55
0 56
0 57
1 58
2 59
3 60
0 61
1 62
2 63
3 64

So, since I have 9001 things to do at the same time, I can't enter zen mode to figure out how to handle those cases, if I just 29.970 them the video gets shorter, if I ignore them it gets longer, someone has an idea? The whole file https://www.mediafire.com/file/b3t1qzl5hoxsmno/CD2VFR.txt/file

[Ceppo]

I woke up to smoke and I got it in 5 minutes. Nailed it! :D

I got what I wanted thanks everyone for the help :)

[real.finder]

I woke up to smoke and I got it in 5 minutes. Nailed it! :D

I got what I wanted thanks everyone for the help :)

can we got the plugin to test with?

[Ceppo]

I actually found a better way. You don't even need DGIndex.

ffmpeg.exe -i "concat:VTS_01_1.VOB|VTS_01_2.VOB|VTS_01_3.VOB|VTS_01_4.VOB" -c copy combined.m2v

Merge the vob files into a m2v.
Then:

ffvideosource("combined.m2v",timecodes="vfr.txt")

The m2v file actually has the timecode file saved into the stream. I noticed it with mediainfo. I guessed it was a timecode and it was.
ffvideosource is x2 faster than d2v import plugins, but has worst seeking if you jump to random frames. The encoder will call them linearly so this is not an issue.

[real.finder]

but has worst seeking if you jump to random frames. The encoder will call them linearly so this is not an issue.

it is issue, if you search in doom9 you will see how bad it is, especially since we don't use the encoder alone as we use some filters and many filters are Temporal so it need frames from earlier and later

if you don't like DGIndex then you have https://github.com/dubhater/D2VWitch

[Ceppo]

Well you get the timecode with ffvideosource and then you encode it D2V source. I used CQTGMC with ffvideosource which calls 7 frames for frames and I had no issue at all.

[real.finder]

even so, don't forget we may use trim! even ffvideosource Developer (myrsloik) would not recommend this, and if you care for the speed it's from new MPEG2 decoder that use multi threads

so with D2VWitch and https://github.com/dwbuiten/d2vsource you should get same speed or even faster since avs d2vsource still use old MPEG2 decoder that single thread

[Ceppo]

I see, I will give it a try then :)

[real.finder]

I see, I will give it a try then :)

also it bad since we need to cut the "vfr.txt" if we did trim, also it will have the same problems of post telecine editing (https://github.com/Asd-g/MPEG2DecPlus/issues/9#issuecomment-1107638503), so what the point aside from unneeded speed (MPEG2 decode already so fast even in very old cpu)?

[Ceppo]

also it bad since we need to cut the "vfr.txt" if we did trim,
That's why python exists, to do that kind of things, it's just a frame shift, I don't think you are going to cut in the middle, and even so it's still possible to script it :)

And about speed, well maybe is unneeded, but it's still an alternative.

[real.finder]

That's why python exists, to do that kind of things, it's just a frame shift, I don't think you are going to cut in the middle, and even so it's still possible to script it :)

And about speed, well maybe is unneeded, but it's still an alternative.

I know that we can cut it (I did in avspmod macro) but that old classic avs way

edit:
I don't think you are going to cut in the middle

there are cases that need cut in the middle :)

[Zetti]

DG have released a test release with flags:

http://rationalqm.us/board/viewtopic.php?p=15680#p15680

[Katie Boundary]

"ignore" should not be used anyway


That's not quite accurate. "Ignore" is a great diagnostic tool. It's also functionally identical to "Honor" if you're dealing with 100% pure NTSC content, and it can be paired with AssumeFPS(24000,1001) if you're dealing with 100% film content. It's also useful if sync with the original audio isn't necessary and you just need raw short clips for music videos, fake trailers, et cetera.