DeComb or TDeint?

[kalorx]

How do I tell which plugin listed in the subject line is better to use for a DVD backup?

I've heard that some DVDs are pure interlaced while other are telecined, and mixed reports as to whether or not both have anything to do with each other.

As can be expected, I'm rather confused. I've been using DeComb for everything with mostly good results, but there are times when the video is jerky despite telecide() and decimate() settings, and it's been suggested that it's because the DVD is "pure interlaced" and TDeint would be better.

Is there a quick and reliable way to know? Does DGIndex report such a thing? Thanks!

[foxyshadis]

MEgui ingludes a function to run a detection algorithm, which will find out whether it's telecined, interlaced, or a hybrid of both, and recommend a filterchain based on that. It's really pretty good at it, too.

To do it on your own, interlaced video appears to move forward every field (every frame combed), where telecined is 3 good frames and 2 combed frames in every cycle of five. There's other forms of pulldown, but that's basic ntsc.

DGindex just reports whatever the mpeg stream says, if it was hard pulldowned (before encoding), it won't be able to get the real telecine order.

[shon3i]

@foxyshadis is that mean that TDeint is better, i see in megui some information about video and i see something like near intrelaced, progressive something called like static frames or something like that.

[Revgen]

TDeint is better used on "Pure Interlaced" content like sports, news broadcasts, game shows, etc.

Decomb is best used on films like Saving Private Ryan or film-based television dramas like Law and Order that use a combination of progressive and interlaced frames in the video.

What kind of video do you have?

[kalorx]

Since MeGUI is out of the question (requiring .NET which wreaks havoc on my PC), I guess I'll just have to get good at determining the proper approach manually.

The video I'm working on is the Disney release of Howl's Moving Castle.

Anyway, this is very interesting information. I've been doing some research into TDeint, so I can't wait for a chance to use it.

[foxyshadis]

AutoGK has a detection algorithm as well, you could take the avs it generates. It isn't as accurate for the more "interesting" sources though.

[Pookie]

Forgive me if this is wrong, but believe the Tools-->Parse D2V section of DGIndex is helpful in determining this

Interlaced:

[GOP]
15 [B]: 15,15...................2
16 [B]: 16,16...................2
17 [I]: 17,17...................2
18 [B]: 18,18...................2
19 [B]: 19,19...................2
20 [P]: 20,20...................2
21 [B]: 21,21...................2
22 [B]: 22,22...................2
23 [P]: 23,23...................2
24 [B]: 24,24...................2
25 [B]: 25,25...................2
26 [P]: 26,26...................2
27 [B]: 27,27...................2
28 [B]: 28,28...................2
29 [P]: 29,29...................2

Progressive:

[GOP]
13 [B]: 13,13...................0
14 [B]: 14,14...................0
15 [I]: 15,15...................0
16 [B]: 16,16...................0
17 [B]: 17,17...................0
18 [P]: 18,18...................0
19 [B]: 19,19...................0
20 [B]: 20,20...................0
21 [P]: 21,21...................0
22 [B]: 22,22...................0
23 [B]: 23,23...................0
24 [P]: 24,24...................0
25 [B]: 25,25...................0
26 [B]: 26,26...................0
27 [P]: 27,27...................0

Telecined:

31 [B]: 39,39...................0 *
32 [I]: 40,40,41................1 *
33 [B]: 41,42...................2 *
34 [B]: 42,43,43................3 *
35 [P]: 44,44...................0 *
36 [B]: 45,45,46................1 *
37 [B]: 46,47...................2 *
38 [P]: 47,48,48................3 *
39 [B]: 49,49...................0 *
40 [B]: 50,50,51................1 *
41 [P]: 51,52...................2 *
42 [B]: 52,53,53................3 *

[Lil' Jer]

The video I'm working on is the Disney release of Howl's Moving Castle.

That should be a progressive DVD. All you have to do is select forced film in DGIndex.

[kalorx]

According to my DGIndex parse of Howl's D2V, it's telecined, which makes next to no sense since DeComb does little but screw up output unless its settings are placed almost off. I had to completely shut off pattern guidance and use low post-processing for it to come out smooth and clean.

Stranger, DGIndex output two different D2V files: one they said that had a "fixed" field order and one labeled as "bad." At first I used the "fixed" one, but the results were just terrible. It was only when I used the "bad" D2V that I was able to get a smooth, clean decimate and telecide.

Howl's Moving Castle has been a frustrating and confusing rip straight from the start given all of this along with its rip protection that I had to figure out how to work around.

[Lil' Jer]

Open up the d2v file in a text editor and scroll down all the way to the bottom and tell me what it reports the Film % at. Considering that other Ghibli release by Disney have been encoded progressive it would be odd that this one wasn't.

[kalorx]

It's at 99.95% FILM.

[foxyshadis]

If you saved the original d2v with force film, you don't need to use ivtc in the script, since that's what force film does. Sounds like that's what happened. If not, it could be some goofy protection and maybe TIVTC would give better results. Still, you should be able to just use force film and not worry about it.

[kalorx]

I always make sure to shut off force film because of some bad experiences in the past, so it couldn't have happened by accident.

But now I have a new problem: I used BeLight (latest as of this post) to output the AC3 into an OGG, but the audio is slightly ahead of the video! I'm positive I have the proper delay, as the "#ms" delay indicator on DGIndex output AC3 reports a 0ms delay.

I used Normal dynamic compression, an output sampling rate of 44Hz, HybridGain mode, 0ms delay, -3db LFE to LR, stereo output, 100% normalize.

The audio comes out slightly ahead of the video! I restored the video and audio to their original lengths for my second encode attempt, so that can't be an issue.

I'd rather not manually mess with the audio delay but figure out exactly what went wrong.

[Lil' Jer]

It's at 99.95% FILM.

Then as I said there is no need to do an IVTC as the DVD is encoded progressive and you can do forced film.

[Lil' Jer]

I used Normal dynamic compression, an output sampling rate of 44Hz, HybridGain mode, 0ms delay, -3db LFE to LR, stereo output, 100% normalize.

Why did you change the sampling rate? Leave it at 48kHz.

The audio comes out slightly ahead of the video! I restored the video and audio to their original lengths for my second encode attempt, so that can't be an issue.

I'd rather not manually mess with the audio delay but figure out exactly what went wrong.

Is it a constant offset or does it get more and more ahead as the video plays? If it's constant you can just use an offset when muxing the audio/video. If it's an ever increasing delay it may have to do with you downsampling the audio.

[kalorx]

It's constant, but by less than a second. I'm afraid to manually offset because they won't be matched perfectly; working in milliseconds is difficult. I'd like to know what went wrong in the first place.

As for downsampling, I always downsample the audio. Don't know why, really, I guess I was convinced it would decrease filesize while keeping quality basically the same.

Anyway, I've checked to see if downsampling was the problem, but it wasn't. The sound is still out of sync even at 48Hz.

[Lil' Jer]

It's constant, but by less than a second. I'm afraid to manually offset because they won't be matched perfectly; working in milliseconds is difficult. I'd like to know what went wrong in the first place.

Why would you be afraid? You can just give the audio an offset when you mux into the container. It might just be that DGIndex gave a wrong offset which I've seen happen from time to time.

As for downsampling, I always downsample the audio. Don't know why, really, I guess I was convinced it would decrease filesize while keeping quality basically the same.

Downsampling the audio from 48kHz to 44.1kHz doesn't give much of any smaller filesize but will it will degrade audio quality. There is no reason to do it.

Anyway, I've checked to see if downsampling was the problem, but it wasn't. The sound is still out of sync even at 48Hz.

Then find out the exact audio delay and pass that to whatever you are using the mux the files into the container. It may take a bit of work, but you can figure it out.

[Lil' Jer]

There's one more thing you can check. Open the .txt file generated by DVD Decrypter when you ripped the DVD. Does it list any offset value for the audio?

[kalorx]

I didn't use DVD Decryptor to rip the DVD. I used DVDFab. I usually use SmartRip, but it couldn't cut through Disney's copy protection.

[manono]

It's a Disney title, so it probably has RipGuard protection. After using DVDFab Decrypter, you should then run it through FixVTS or VobBlanker to remove unreferenced cells. This is perhaps (probably?) the cause of your audio asynch.

As stated before, just Force Film the thing and neither IVTC nor deinterlace it.

Pookie's earlier post was well intentioned but wrong.

Edit: Oh, and I'd be real careful if I were you, kalorx, about denigrating programs you don't know how to use (Decomb) or whose information you don't know how to interpret (DGindex/Parse D2V).

[Lil' Jer]

It's a Disney title, so it probably has RipGuard protection.

Which is why I'm glad I have the R2 Japanese version. That and the 6.1 DTS-ES track. :p

After using DVDFab Decrypter, you should then run it through FixVTS or VobBlanker to remove unreferenced cells. This is perhaps (probably?) the cause of your audio asynch.

Ahh didn't even think of it having ripguard on it. That'd probably fix the problem.

[neutrogenik]

I got a question about Decomb : i have a video that has progressive and interlaced frames, that DGIndex.exe detect well.

I was wondering if the filter can use the .d2v infos to know when the frame is interlaced or progressive, and if yes what is the script to do that ?

[Lil' Jer]

I got a question about Decomb : i have a video that has progressive and interlaced frames, that DGIndex.exe detect well.

If by detect you mean it reads the flags in the mpeg-2 stream then, yes.

I was wondering if the filter can use the .d2v infos to know when the frame is interlaced or progressive, and if yes what is the script to do that ?

You can use the IsCombed() command to detect if their is combing. That's all I can think of.

BTW is this hybrid material or just contains progressive and interlaced frames due to being telecined?

[neutrogenik]

If by telecined you mean recorded from TV source then yes it is :)

[Lil' Jer]

If by telecined you mean recorded from TV source then yes it is :)

Well it could mean that or not. So is the source hybrid or telecined? The point is that DGIndex doesn't "detect" anything as it just reads the flags in the mpeg-2 stream and many times the flagging can be flat out wrong.

[neutrogenik]

In fact i don't know what does "hybrid" or "telecined" stand for so i give you a sample so you can maybe tell me :D

http://rapidshare.de/files/17046879/original.m2v.html

[Lil' Jer]

In fact i don't know what does "hybrid" or "telecined" stand for so i give you a sample so you can maybe tell me :D

This would be a good read then: http://www.doom9.org/ivtc-tut.htm

[Lil' Jer]

Okay, first of all you didn't mention it was PAL. Secondly, I stepped through your clip and didn't see a single progressive frame. Looks like this would have to be deinterlaced.

[neutrogenik]

Thanks 4 the link :)

Well in fact that some frames were progressive because the interlaced artefact are hard to detect but if you say that all the frames are interlaced i should maybe just use Tdeint() right ? :)

[Lil' Jer]

Well in fact that some frames were progressive because the interlaced artefact are hard to detect but if you say that all the frames are interlaced i should maybe just use Tdeint() right ? :)

Well the part with the red lighting looked progressive but even still it looks like you will have to deinterlace. I tried to check if it was just field-shifted and a frame matcher could be used, but that failed. Tdeint would definitely be your best bet from what I can tell.

[neutrogenik]

In fact i don't understand why are the PAL video interlaced because if the PAL telecine is only made by taking a frame and separating it into 2 fields and then playback by putting the 2 fields together the frames should be perfect :??:

Edit : maybe the "sécam" mode that is used is france is for something in this ?

I'm just trying to understand how is made the source video to desinterlace it better.