Cadence problem with JES

[Eugenia Loli-Queru]

I shot a clip with my HV20 at 24f, I export it with imovie using the Apple Intermediate Codec, it creates a big .mov file, I open it in JES in order to do the 3:2 pulldown using "use top frame" and "detect cadence", and unfortunately, right after the new fixed video is created, there is an error:

Cadence Phase not found err ERROR # 5017

Any ideas what's going on? Even removing "detect cadence" creates the same error.

[Eugenia Loli-Queru]

Guys, when you use JES Deinterlacer, do you use the m2t files to do the 3:2 pulldown, or the AIC codec in .mov format?

[icpark]

I got this error too. If anyone can help, that would be amazing.

-Ian

[Eugenia Loli-Queru]

I think the problem is that JES can't remove the pulldown with .mov AIC files, but it can with .m2t files. So, you need to capture with an application that saves down .m2t files during capture, not iMovie that by default saves down .mov AIC files.

[directore]

not my experience, JES-DE works only with *.mov files, not *.m2t. In author's own words (his reply to my question):

>> Three suggestions, neither of any significant priority:

a) your s/w doesn't read transport files (.m2t) which is how some people capture stuff from HV20. In this case one has to use MPEG Streamclip to convert it into .mov. Avoiding that would be nice.
>>

I know but I've decided to keep things simple and rely on QuickTime for import of non-native file types. Unfortunately QuickTime support for MPEG2 has always been very limited. That's one of the first things I would change if I got Steve Jobs' job.

[daci]

JES can convert AIC movie, or any Quicktime movie your system can handle. Unfortunately it cannot process m2t directly which would've been great. I'm not sure about the error though.

[Eugenia Loli-Queru]

Well, it did not work in my case, neither for this guy's case. The clip I had used was captured via iMovie and it was 24f, saved down as .mov in AIC format (as iMovie does by default). JES would give me that error. I emailed the creator back then and uploaded a 80 MB file for him to try. He was replying some weird stuff to me:

>> here is the clip that wouldn't work to remove the 3:2 pulldown. It is
> >shot at 24f inside a 1080i stream using the Canon HV20, saved with
> >iMovie 05, while the latest version of Quicktime is installed.

This is a normally interlaced movie. You cannot/need not inverse telecine it. Your camera probably has more than one mode. This clip was shot in normal interlaced mode (60i). Just look at the images. Each frame is interlaced. If this was telecined video only 2 out of 5 frames would have interlace in them.

Even after I explained to him what 24f was and how Canon saves that video inside a 60i stream, he wouldn't get it. So, I honestly don't know how people got pulldown to work with his utility.

[directore]

He makes sense, you don't seem too.

[Eugenia Loli-Queru]

How am I not so? AFAIK, 24f is a stream inside 60i. When you do the pulldown, you remove the extra frames. The video I sent him WAS SHOT at 24f.

[directore]

look if the clip you sent him shows interlacing artifacts in every single frame of it than that clip is NOT a telecined 24p footage, is something else, and what precisely it is depends on how you generated it. But that's entirely irrelevant here, no 2/3 interlaced/non-interlaced telecine cadence present, no way it can be removed by anything, JES included.

[Eugenia Loli-Queru]

The 24f footages on Canon HV20 cameras are ALWAYS interlaced. This is why we transform them from 24f to 24p. The clip was captured by iMovie and saved down as .mov Apple Intermediate Codec. When playing back that file on Quicktime, it is 30fps, interlaced (as it should report it). But it was shot as 24f, and so after removing pulldown, it SHOULD transform that file to 24p.

Now, as you said, it might be missing cadence information, which is another issue. Point is, the clip was unaltered, copied directly from iMovie's folder right after capture. If iMovie messed up with it, as I said, it's another issue altogether.

[Ian-T]

Someone answer me this...the 24P wrapped in a 60i timeline from the HV20....doesn't that just simply mean that 24 frames are progressive and every other (6 total) are interlaced? All together they would produce 29...fps? Reverse telecine then would be just removing the interlaced frames right?

If so, then directore would be correct to say that if all frames have interlacing artifacts then the original footage must have simply been shot in 60i.....and not in 24p 2:3 right?

Edit: I mean you're both "sort of" correct but Euginia I'm not sure if the 24p is "all" interlaced (hence "P"). Please somebody correct me if I'm wrong....but I'm learning like the rest of us.

[directore]

instead of thinking aloud here I would humbly suggest you go and read for example this: http://en.wikipedia.org/wiki/Telecine

[Eugenia Loli-Queru]

They are ALL interlaced. All of them. When you look at the bit stream, it's really 60i. If you assume that your stream will be handled as 30f and you have 2:3:3:2 24f (but not 3:2 24f), then there are 24 frames out of 30 in the 30f frames that match the 24 frames you're interested in (though not at the highest quality). But the actual data is always 60i. There may (or may not, as in our HV20 case) be some attributes that tell you how the fields are related, but there are always 60 interlaced fields.

Applications like Cineform had to develop a special case in their algorithm to make frame comparison and remove the duplicated ones. If it was a bastardized 24p and 5.9i, then no one would be shouting at Canon to include attributes in the stream to help apps to do the pulldown. All it would need to do, would be to remove the interlaced ones. But that's not the case, because all frames are interlaced, and without attributes. And so the app that removes pulldown must be extra careful with the HV20.

Directore: We are talking about removing pulldown here, and I am pretty experienced with the HV20 case. Now, other cases of "telecine" might be different and you might be reffering to them instead of the HV20 one. HDV cameras --by standard-- only shoot interlaced. And so Canon had to *hack* the standard in order to record in 24f -- which is NOT real 24p. Which is why we need to remove pulldown (and when you do that the interlacing gets removed too), in order to TRANSFORM it to 24p. But the original capture is NOT 24p.

[Ian-T]

... But the original capture is NOT 24p.Ok. When you say this you mean when we "capture" the footage off the tape it is not 24p but interlaced. I understood that. The "standard" interlaced format you are referring to is what went to-tape but the sensors in the HV20 from what I understand is progressive. It just lays it out on tape as an interlaced file. To me it's sort of like the Hollywood DVD's we watch...they were originally filmed as 24 fps on film but when transfered to DVD the footage was telecined per industry standard. DVD players with progressive scan capability automatically reverse telecine the footage for our viewing pleasure. It is still real 24p we are viewing from the DVD right? So if the HV20 orginally has progressive sensors but saves it as 60i then wouldn't the "transformed" part you mentioned really have taken place FROM 24p TO 60i and all we are doing is reversing that process by removing pulldown?

I guess the bottom line of what I'm asking is ....aren't we just reversing the orginal transformation? I hope I didn't confuse you but I'm just trying to get a grip on this whole process.

[Ian-T]

Oh...and thanks for the link directore. I pretty much always understood that whole process. I was just thinking in terms of full frames and not half fields.

If the above thought (about HV20 having progressive sensors) is true then my thinking was when it is telecined....those newly added fields are interlaced (not the original progressive frames from the progressive sensors).

[Eugenia Loli-Queru]

>aren't we just reversing the original transformation?

Yes. But unfortunately, each manufacturer uses its own tricks and standards when they save down the 24f transformation to tape. So, an app like JES or Cineform, must test with all these cameras and figure out ways to do the re-transformation the right way. For example, Canon's pro cameras have attributes to tell the app which frames should be removed, while Sony uses a different cadence. The HV20 uses a yet different one, and without any attribute guidance. In other words, it becomes messy for apps to properly support and remove the right frames for each camera, because there is no standard on how to do that for HDV as it is for DVDs. Each manufacturer re-invents the wheel each time they release a new camera. DVD Players can re-transform to progressive, because there is a predefined, hard standard on how to do it. For HV20, we need special support from each app, and from what I know, the only company who has actually ordered an HV20 and specifically tested with it, is Cineform. All the other solutions are simply don't work always or well (e.g. Premiere LE, JES), or the developers must at least have access with unmodified 24f .m2t files from the Canon HV20 directly, in order to test and figure out what to do with it in their app.

For example, I am not convinced that the trick/hack explained (and I personally also use) in the sticky thread on this sub-forum with avisynth/dindex/virtualdubmod creates perfect 24p as it should be (for example, some audio sync problems exist on some PCs which means that some frames were removed by the script when they shouldn't have).

And so it gets really hard trying to explain this to a developer, like I tried to for JES, and he still didn't "get" it.

[Ian-T]

I hear you. But even from the definition of how pulldown is applied to footage.....I think that it is ...universal. What I'm saying is 3:2 pulldown is 3:2 pulldown. The HV20 apples it just the same way the DVX does with only one exception (which we all know) and that is with the absence of flags. It does not add the flags to its footage. This was done on-purpose by Canon (IMHO). But flags are not the only way to reverse this process.....it's just that...up until the HV20....this was the norm and expected.

You have to remember when the DVX first came out...there was also...NO support to remove its pulldown. Then all of a sudden we are using the flags to remove its pulldown. I firmly believe that flags are not the only way based on this fact (obviously as you mentioned there is Cineform which does an excellent job wothout the flags present).

I really believe that the HV20Pulldown.exe process is removing the proper extra frames. I used it along with Cineform and ran them side by side in V-Dub...no discernable difference...same exact frames removed...and no ghosting. I thinnk the software (or process) is able to see the original progressive frames within the 60i fields and is able to put then back together intelligently while discerning the extra fields thrown in there to give it the total 29.970 frames. I might be totally off but that's my thinking anyways.:hv20-smilie03:

If the above is the case...then that's why I orignally thought those extra frames...when put together...are interlaced. Think about it....the software reassembles the original 24 frames....sees the 6 newer interlaced frames....kicks them to the curb....WALA....true 24p. Just a theory however.