hybrid source with 2% NTSC / decomb questions

[wfn1]

I've been trying to transcode(CCE) Fellini's 8 1/2 (Criterion edition) and since it's my first experience with a hybrid source I thought I'd ask the omniknowledgeable forum members a few questions.

1) When encoding a hybrid source with relatively low action, at what % of NTSC should I start considering decomb.dll? (in my example it was 98%film 2%ntsc for the actual movie) Since I thought the transfer was not perfect to begin with I frameserved it with DVD2AVI with "no field order" and ran it through Telecide(post=false) Decimate(mode=1,threshold=50) in my .avs script.

2) When do I want to use Telecide() vs. Telecide(post=false), I've been told that there is no rule of thumb but supposedly post=false works better on cleaner/more modern transfers.

3) Is it normal for the end result .mpv to come out with drop_frame flag set? I had to do a -nopulldown -drop_frame false to import chapters in Maestro as it does not like the drop for that. (later reverted back to dropped for subtitle sync)

Sorry, if this belongs in a newbie forum and thank you for any feedback.

[JohnMK]

Originally posted by wfn1
I've been trying to transcode(CCE) Fellini's 8 1/2 (Criterion edition) and since it's my first experience with a hybrid source I thought I'd ask the omniknowledgeable forum members a few questions.

1) When encoding a hybrid source with relatively low action, at what % of NTSC should I start considering decomb.dll? (in my example it was 98%film 2%ntsc for the actual movie) Since I thought the transfer was not perfect to begin with I frameserved it with DVD2AVI with "no field order" and ran it through Telecide(post=false) Decimate(mode=1,threshold=50) in my .avs script.


No Field Operation, you mean. You need to use mode=0 or 2, not 1. 1 leaves it at 29.970. I believe something like this might be better:

Telecide(guide=1,post=false)
Decimate(mode=2) or (mode=0)

Neuron2, however, cautioned me against using mode=2, but his help file in decomb says this is better for IVTC of 3:2 pulldown material. Neuron2, care to clarify?



2) When do I want to use Telecide() vs. Telecide(post=false), I've been told that there is no rule of thumb but supposedly post=false works better on cleaner/more modern transfers.


For most DVD hollywood feature movies you shouldn't need to post process, but I do just in case anyway. It doesn't hurt. Yeah it slows it down a little bit, but oh well, CPUs are fast. But you really shouldn't need to use decomb on this movie. DVD2AVI Forced FILM can be trusted at this percent FILM.


3) Is it normal for the end result .mpv to come out with drop_frame flag set? I had to do a -nopulldown -drop_frame false to import chapters in Maestro as it does not like the drop for that. (later reverted back to dropped for subtitle sync)

Sorry, if this belongs in a newbie forum and thank you for any feedback. [/B]

Can't answer your last question, sorry. :)

[wfn1]

Thanks for your reply, JohnMK!

I believe something like this might be better:

Telecide(guide=1,post=false)
Decimate(mode=2) or (mode=0)


So, a script like that should leave me with 24fps FILM if I understood you correctly? That would actually be quite nice and I would not have to waste the unnecessary bandwidth in CCE. I'm afraid to use mode=2 though, just in case it cuts out something useful.

Would you say that mode=1 would be the most safe/almost loseless option for hybrid sources? Or am I just full of sh*t?

Also, if you happen to know if there is a quick way to decide between
"mode=0", "mode=2" and "mode=1, threshold=50" by stepping through the fragment in question in VDub with telecide(post=false) like it suggests in IVTC guide?

I would really like to hear what Neuron2 has to say about it, after all if a certain way is better for hybrid than what's in Doom9's decomb.dll guide, might as well amend it or offer it as another option.

[Richard Berg]

I don't do DVD rips so I can't make a direct comparison, but one of the TV shows I'm capping right now (Andromeda) has about the same ratio (5% video source / 95% film source). I decided to go with full-out Telecide(guide=1).Decimate(mode=2) -- terrific output on the film material, and while the CGI sections obviously get choppy they're no worse than any other method I tried short of going to 60fps.

[wfn1]

hmm, that's what i'm trying to avoid at all costs, I want the copy to be as close to the original as possible without having to split to 2 DVD-Rs. So, I guess I'll play more with mode=0 or mode=1, threshold setting. Thanks for your input!

[JohnMK]

Actually in my experience mode=1 on FILM material is a bad idea. Telecide spits out progressive frames + one duplicate in a series of 5. Mode=1 tries its best to make that duplicate a blend of the two surrounding progressive frames, but doesn't seem to ameliorate the problem altogether.

[manono]

Hi-

Listen to JohnMK. You don't have hybrid material. So, don't use Mode=1. It's just a movie and shouldn't have any true interlaced video in there. Telecide().Decimate() should work fine.

[wfn1]

by that logic i can just serve it as 'force FILM' in DVD2AVI though...

[DJ Bobo]

That's it, just use "Force Film" in DVD2AVI.

[JohnMK]

Originally posted by manono
Hi-

Listen to JohnMK. You don't have hybrid material. So, don't use Mode=1. It's just a movie and shouldn't have any true interlaced video in there. Telecide().Decimate() should work fine.

As DJ Bobo states, just use Forced FILM.

Now, for something completely off-topic, what's the meaning of this contained within the DecombHelp.html file?

About mode=2:

"Note that for the most reliable IVTC of 3:2 pulldown material, you should use pattern guidance together with this mode of decimation."

Exactly how necessary is this? From what I've seen, neuron2 is telling us to avoid mode=2 except for anime.

[wfn1]

please share your wisdom on the:


Also, if you happen to know if there is a quick way to decide between "mode=0", "mode=2" and "mode=1, threshold=50" by stepping through the fragment in question in VDub with telecide(post=false) like it suggests in IVTC guide?


subject. Thanks!

[Asmodian]

If you notice some interlacing after the force film (the 2% that didn't have the pulldown flags) then you can use decomb to IVTC but if there is no visible interlacing then force film will encode a bit faster. I noticed some interlacing on some Sherlock Holmes DVD's when using force film when the % film was around 98%, while decomb gave me perfect progressive video (but I had post=true so it might have cleaned up a few interlaced frames, I didn't look through it that closely).
just my 2 cents

[LB]

Try it, the ivtc filter that is. It doesn't work on all sources. It's strange. If decomb or tmpg work, then usually IT wont, whereas if decomb or tmpg don't ivtc the source well, then IT usually works. There is nothing to configure except whether the source is top or bottom field order. But I tried EVERYTHING on a few Region2 dvds I have and NOTHING worked to remove the interlacing on certain hybrid parts, BUT when I tried "IT" 95% of the interlaced frames were removed, NO blending! Give it a shot, it's definatly worth trying out. Search the threads for like IT and IVTC and ye shall find.

LB

[JohnMK]

I don't think IT would make a good keyword. It falls below the minimum length and it's (!) a very common word.

[Wilbert]

See our FAQ for download (Q28). We are still missing a proper documentation about this plugin.

[manono]

Hi-

by that logic i can just serve it as 'force FILM' in DVD2AVI though...

Your's is faulty logic. A full strength IVTC will work on any material (more or less) that was telecined in the first place. Force Film will only work perfectly on telecined material where all the flags have been set properly. Of all the telecined material that I see, about 90% or so must be IVTC'd, and if only Force Filmed will leave behind at least some interlacing or jerkiness (and at percentages up to 99%). The only time that Force Film is guaranteed to work is when it shows as Film (100% Film). The guides say to Force Film when the Film percentage is equal to or greater than 95%. That may work, but isn't guaranteed.

...quick way to decide between "mode=0", "mode=2" and "mode=1, threshold=50"

For the purposes of your movie, you can forget Mode=2, as it's mainly for anime. Make your .d2v with no Field Operation, open it in GKnot, and go to the places that show in DVD2AVI as interlaced or NTSC. Step through it frame by frame and if you see interlacing in every frame, then it's interlaced and Mode=1 may be the answer if you can't live with the jerkiness resulting from bringing it down to 23.976fps. If it shows the usual 2 interlaced frames in every 5, then IVTC it. Or doing it your way, make the .avs with Telecide(Post=False) and no Decimate(5). Go to those same places, and then you're looking for no interlacing and one duplicate frame in every 5.

[wfn1]

thanks for your help, manono!


If it shows the usual 2 interlaced frames in every 5, then IVTC it

Should I IVTC it in DVD2AVI or would decimate(mode=0) would be better?

[manono]

Hi wfn1-

You're fast on the reply-you must want to get this (great) movie finished.:) Ordinarily, when speaking of IVTC, we are referring to the "full strength" solutions such as Decomb, GreedyHMA, IVTC2.2, or IT. Force Film in DVD2AVI is a somewhat "weaker" solution that only follows the flags, and doesn't use the "brute force" methods of field matching or pattern detecting.

But to answer the question-to be absolutely safe, I'd use Decomb if I were you. Force Film may work, but I've been burned too many times by using Force Film and then finding interlaced frames afterwards. Force Film is faster (particularly if the deinterlacer is turned on in Telecide()), but Decomb is surer.

[wfn1]

so in other words if i want to IVTC it I would be better off with

Telecide(guide=1,post=false)
Decimate(mode=0)

rather than DVD2AVI's force film.

btw, I made 2 copies of the movie, one with

Telecide(post=false)
Decimate(mode=1,threshold=50)

lost some bandwidth on the encoding part since it's 29.97 fps.

and one with just force FILM in DVD2AVI, I'll watch both tonight and let you know which one turned out better. I will also try to use decomb to IVTC and see if the end result is better then DVD2AVI's force FILM.

Cheers!

[Guest]

@JohnMK

Mode 0 vs 2: If in a cycle you have multiple different strings of duplicates, e.g., n d d n d, where n is a new frame and d is a duplicate, mode 0 will take the frame that is most similar to its predecessor and not the one that is part of the longest string of duplicates. For normal films, these sequences are rare and it won't matter; use mode=0. For animations that are animated by duplicating or triplicating (or worse) each frame, you need to decimate from the longest duplicate sequence to avoid causing jerkiness; use mode=2.

I need to correct the statement in my help file that you referred to. Thank you for pointing it out.

@wfn1

You might also consider using Forced Film but passing the result through FieldDeinterlace(full=false). I know my hero manono is holding his hands over his ears and grimacing, but you may find it satisfactory for your purposes.

[wfn1]

neuron2,

thanks, i'll try fielddeinterlace(full=false). Also, any idea as to why the resulting .mpv (encoded 29.97 fps, Telecide(post=false) Decimate(mode=1,threshold=50)) has a drop_frame flag set?

manono,

so, do you think :

Telecide(guide=1,post=false)
Decimate(mode=0)

will be an efficient method for IVTC'ing?

[manono]

Hi-

Up to you-there shouldn't be much difference between Force Film with Fielddeinterlace(full=false) afterwards, and:

Telecide(guide=1,post=false)
Decimate(mode=0)

I'd check it real carefully before encoding if I chanced doing it that way though, to make sure no interlacing slipped by Telecide(). Or:

Telecide(guide=1)
Decimate(mode=0). Or:

Telecide(guide=1,Blend=false)
Decimate(mode=0), in case you prefer interpolation over blending. Maybe do a chapter first, or use the Trim command and do a few thousand frames to see if you notice any differences.

[wfn1]

Thank you Manono! I'll try a couple of things as soon as I resolve one other problem I'm having with max bitrate being uncooperative, since it would be offtopic here I asked my questions in authoring.

my max bitrate problem :( (http://forum.doom9.org/showthread.php?s=&threadid=36863)

[JohnMK]

Originally posted by Asmodian
I noticed some interlacing on some Sherlock Holmes DVD's

Which Sherlock Holmes DVDs? The ones with Jeremy Berret as Sherlock Holmes? Is this the boxed set?

[wfn1]

manono,

also, I wanted to ask you if I need to apply decomb in both CCE passes, i.e. i'm using the robshot method of doing 6K CBR encode to build the .vaf file and then I'm doing a single pass of VBR encoding. In this case, do I need to run decomb.dll during both passes or do I only do it for the 2nd run VBR encoding. I've been decombing both time so far.

[wfn1]

Just finished doing 8 1/2, my original movie that made me start this thread. Here are my findings:

DVD2AVI force FILM produces jerky output and some interlacing(very little) is present in the resulting movie.

DVD2AVI force FILM + FieldDeinterlace(full=false) produces very jerky output but the interlacing is gone.

Telecide(guide=1)Decimate(mode=0) as well as any mode 0 modification like post=false or blend=false produces jerky output in certain scenes, no interlacing. Essentially during playback if you're watching at least for 2-3 minutes you will notice 3-4 hicckups like if the frame were omitted.

Telecide()Decimate(mode=1,threshold=50) produces best results and the quality degradation from encoding 29.97 fps source is almost non existant.

At the moment I am working on Bunuel's "That Obscure Object of Desire" Criterion Edition and it pretty much exhibits identical behavior to 8 1/2 however it has a slightly higher NTSC %. DVD2AVI reported 6% NTSC when saving the .d2v. I will be encoding it in Telecide()Decimate(mode=1,threshold=50) as it seemes to have produced the best results after 4 hours of experimenting with different fragments of the movie.

I would also like to note that previewing either title in VirtualDub with Telecide(post=false) (as suggested in IVTC guide) reveals no visible interlacing.

What do you make of all this Manono, Neuron2, JohnMK?

[JohnMK]

Which container format are you using. Ogm or avi?

[wfn1]

what do you mean by container format?

I just rip in file mode in DVDDecrypter, frameserve with DVD2AVI and use avisynth 2.06 as a wrapper to feed into CCE 2.5.

[manono]

Hi wfn1-

As I was reading your post I thought to myself that I was wrong and there really might be some pure interlaced parts in there. But then when you said that previewing either title in VirtualDub with Telecide(post=false) (as suggested in IVTC guide) reveals no visible interlacing, then I concluded I was correct with my original assesssment (without having seen the DVD). Assuming you were checking the parts that played jerky, they should show interlacing if, indeed, it's a true hybrid with some interlaced sections. So, my guess is that something else is at work here.

Do you have the latest Decomb version? There have been improvements since the earlier versions. Have you tried the Quality=3 setting in the Decimate line? Because as near as I can tell, if what you're saying is correct, then the wrong frame is being chosen as the duplicate and being dropped. Perhaps it's a hardware issue. I gather you're making DVDs. Are they jerky when played on a standalone unit, or are you playing them on a computer? If on the computer, are these panning and scrolling shots, and if so, is your computer powerful enough to handle them (no insult intended-just checking)? Maybe neuron2 would be interested in having a look at one of the problem sections, but I can't speak for him.:) If worse comes to worst, you could give IT (as LB mentioned earlier) or IVTC2.2 a try.

And yes, you need Decomb in all of the CCE passes. The frame reconstruction and decimation information isn't saved anywhere, so it gets done anew with each pass.

You're a man after my own heart-those are some very fine classic movies, but sometimes the telecining isn't up to par (although Criterion should know better). But it's not like doing a current Hollywood movie where in most cases you can Force Film and be done with it. But all of your work will pay off in the end with a backup you can be proud of for many years. If you think those movies are tough, try playing with some of the silent films I'm working with these days. Weird frame rates, bad telecining, crop values that change constantly, frequently blended fields, noisy source material requiring testing out various filters to make them more compressible without ruining them, all combine to make for a very challenging job.

[wfn1]

I apologize, I was doing That Obscure Object of Desire late at night and I must have messed up with my VDub assessment. I re-checked it tonight and I did notice interlacing in certain parts, it's not as pronounced as it becomes when you Force FILM but you do see the jagged edges.

Do you have the latest Decomb version?

Yes, I'm using the 4.00 without the latest PAL addition.

Have you tried the Quality=3 setting in the Decimate line? Because as near as I can tell, if what you're saying is correct, then the wrong frame is being chosen as the duplicate and being dropped. Perhaps it's a hardware issue. I gather you're making DVDs. Are they jerky when played on a standalone unit, or are you playing them on a computer?

I have not tried the Quality=3 setting, it will be my next thing to try. So far I've been very satisfied with Telecide()Decimate(mode1,threshold=50) output but I would not mind to use some extra bandwidth the 24FPS FILM provides seeing as how most of my favorite movies run in excess of 2 hrs.

I have tried previewing with both PowerDVD-XP and on my standalones (Pioneer DV-525, Pioneer DV-C503) the effects are the same. My machine should be able to pull it off anyway, AMD 2100+,512 DDR, Matrox G400, Seagate Cheetah 73LP as a project drive.

I'm also going to be doing most of US released Antonioni, Godard & Renoir films, so it should be fun. As far as silent movies go I have some Chaplin's short films that should present some sort of a challenge as I've heard they're 16-18fps...

Thank you so much for the CCE tip. Also, a quick question have you heard anything about this issue with ScenChap 1.27 (http://forum.doom9.org/showthread.php?s=&threadid=37179) ?

[wfn1]

unfortunately quality=3 in mode=0 does not do any good for me because still an invalid frame is being deleted.

[wfn1]

well, i've tried all decimate settings and i also played with IVTC 2.2 and cannot achieve a good result with Antonioni's L'Avventura Criterion Ed. Mode=0 or Cycle=5 or Mode=2 with quality=3 all end up taking out some unnecessary frame. Maybe the pattern is too complex, because with those settings I'm getting a jerk in uneven intervals but at least 2-3 times a minute. So I broke down and started encoding it 30fps non-progressive. Until I find something that works I guess that's how I'm gonna have to work this out.