Create interlaced (with different fields) from progressive footage?

[l93812963]

I have video footage that was originally 23.976p converted to 59.94i (with telecine) and then converted to 29.97p. The only footage I have is this 29.97p video.

Is there a way to go backwards one step, converting the 29.97p video back to 59.94i with different fields within the frames.

Through all my tests, exporting as interlaced -- the two fields are the same, taking them from the progressive frame. Is there a way I can shift/slide a field later, so the frames are all messed up with different fields? --So lower field of frame 1 and upper field of frame 2 now makeup frame 1. Lower field of frame 2 and upper field of frame 3 now make up frame 2. And so on.

The reason why I want this is a super long story. Getting the original source is not an option.

I have access to Premiere, After Effects and Avid Media Composer. Thanks.

[l93812963]

I'm realizing now that I should probably explain the backstory. Because I think I'm able to get the interlacing back with Re:Vision FieldsKit Reinterlacer plugin, but it's not really solving my problem. Maybe you guys can just solve my problem.

We do a localization of a foreign animation series. The series used to be delivered to us on HDCAM tape (59.94i). We would localize it and output in the same format at 59.94i. This past season, they were sending us digital video files of the episode at 29.97p. That was fine, because that helped our workflow. Editing and rotoscoping was much easier without having to deal with the interlacing.

We assume they animate at 23.976 and add pulldown to get it to 59.94i or 29.97p, based on that we can see the hold frames on every fifth frame. It also stutters a little, which is apparent on panning shots. Now this has always caused problems when converting for PAL countries at 25fps. But this time, a broadcaster rejected the 25fps 50i conversions our digital lab was making. They were fine with past seasons, but not this time. I saw them and they were terrible. They said they tried so many different way ...and they're supposedly the experts.

So we decided to see if we can fix it in-house and do the conversion ourselves. Which means doing a reverse telecine and bringing it back to 23.976p and then convert that to 25fps 50i via speed change. The cadence of the pulldown is all over the place and changes almost every shot, so it's not easy. We've tried software like Compressor that claims to detect changing cadence, but it didn't work well. After Effects seems to have a hard time guessing the pulldown on this footage during my tests. I even tried some plugin in Avidemux. Didn't work throughout.

My solution based on what would be the easiest workflow for my editors and myself is to render out the whole episode 5 times, with an offset of 1 frame per render. Bring this into Premiere in a 23.976p sequence. It's already pulling a frame every five frames. So it's basically doing a blind pulldown removal. Now because each render is offset by one, at least one of the video renders will have the correct shot where there are no hold frames. And we re-assemble the episode that way, shot by shot. Very tedious, but the best solution I have right now.

My boss just yesterday had a question about why were previous seasons accepted by the same broadcaster. So he asked me to do a test with the old episodes at 59.94i and do the same shot swapping I'm doing now to remove the pulldown and see if it's any different. In my mind, it would just make things worse because it's interlaced. But holy moly, Premiere did a way better job at removing the pulldown this way when I dropped it in the 23.976p sequence. Whereas when I put in the current the 5 renders (offset by 1 frame) of the 29.97p footage into a Premiere 23.976p sequence, only 1-2 of the renders would have the correct shot. Now when I put in 5 renders (offset by 1 frame) of the old 59.94i footage into a Premiere 23.976p sequence, 4 of the render are correct. That's a lot more chances of it looking good. That's probably why it was accepted by the broadcaster before.

So now my goal is to see if I magically get these 29.97p videos back to 59.94i for Premiere to do whatever the heck it's doing to make my life easier. I feel like I'm killing myself now. After my editors do the shot swapping assembly, I'm checking it and sometimes I'm redoing every one of their shots. I'm not sure why they aren't getting it right. But if Premiere can make it so maybe 80% of the video is already good, I think this too shall pass.

Or if someone can just tell me what magical software will just do this for me with a press of a button. Seems like there should be something and I think a computer can do it. But the things I tried just don't work. Maybe because it's animation. And sometimes there's not much movement in the shots and it doesn't know what to do, so it fails.

[l93812963]

I'm sure there is a way to do the same thing we're doing in Premiere and do it in After Effects instead and not have to render the episodes out 5 times each. But we needed a solution fast and get a bulk of the episodes quickly. There was no time for my Avid editors to learn After Effects.

[rodneyb56060189]

gotta simplify this....

your source stuff is 29.976P, right ? And you want to export for PAL ?? That would be like 25 ? I don't do this stuff so I'm kinda guessing.

Let's say 30fps Progressive, for simplicity, and 25 FPS interlaced OK ?? Is that what you want ??

Can't you take one clip of the 30 stuff and put it into a timeline of matching specs and export custom PAL and choose the opposite field order than is normal ?  I forget what they are ( top first or bottom first ). ???  Is that what you're trying to do ??

duh.

[l93812963]

Oh yeah, I never said why it looks bad if you just did what you suggested. If you did this in Premiere, the hold frames (pulldown) looks more apparent and there is bad jutter/jerkiness (mostly apparent on panning shots). Some programs like Avid Media Composer will make blended frames. I'm sure you can make each other's software do the same with some settings. But either way you'll have blended frames or bad jutter or both, which I've seen in the conversions our digital lab made.

But I think I just solved my solution! In next comment below...

[l93812963]

Okay, I think I got what I need to resolved my situation --to get my footage back to 59.94i properly so that Premiere looks at it and brings it back to 23.976p and removes the pulldown correctly for most (80%ish) of the shots.

I used Re:Vision FieldsKit Reinterlacer. Tried so many settings in Premiere and After Effects. I'm pretty drunk right now. But tripled check and I think I got it to work.

Use Re:Vision FieldsKit Reinterlacer in After Effects and apply it to the video.

Field Order: Upper Field

Output Type: Keep Frame rate

Export in Media Encoder: Quicktime ProRes (thank you Adobe for bringing ProRes to Windows), 59.94, Upper Field first.

...our native format was Avid DNxHD, but I couldn't any DNxHD preset to get it to true 59.94. I tried the 59.94 presets, but looks like it was outputting 29.97. And that's how Premiere was looking at it after I exported and imported it. It saw the export as 29.97. I know. I'm still kinda confused on 59.94i ...if it's really 59.94 frames interlaced or 29.97, but 59.94 interlaced fields. All this is confusing. But after the trial and errors. This is what worked for me. So don't blame me!

Again, I rendered out 5 videos of it, offsetting by one 1 frame. Dropped them into a Premiere 23.976 sequence. Now all of them have the pulldown removed correctly for almost all of the shots.

This is insane. I think I deserve a medal ...for a problem that doesn't really exist anywhere else. Really, the original company should have produced this at their native animation frame rate (23.976) and gave us that. And for their local broadcast, then do their standards conversion to 59.94i afterwards.

Thank you for listening.

[l93812963]

Okay wait. The thing I did last night didn't work. Turns out it sped up the footage and I didn't notice that last night when I was stepping through the shots. Also I was drunk and delirious. I feel like this is Groundhound Day where I'm waking up to the same problem every day. But it's the Daylight Savings version.

[rodneyb56060189]

yeah, I was gonna mention the blended frame thing, but with extra frames it would probably just look like blurry extra frames.

hehe.. pretty weird.

[chrisw44157881]

next time just count the wwssw or whole or scanline frames and then use the AE pulldown type that matches what you see when going 5 frames forward. now if the cadence changed on every shot, you'd have to splice them up or use terenax to detect changing cadence. there used to be a virtualdub plugin that auto detected cadence changes somewhere. also, it may not have even been a standard pulldown. some older cameras used 2232 or 2323 pulldown just to keep things interesting.

[rodneyb56060189]

yikes. The plot thickens.

Here is what you said in past … between the lines of ====

=================

I have video footage that was originally 23.976p converted to 59.94i (with telecine) and then converted to 29.97p. The only footage I have is this 29.97p video.

==============

==============

Now this has always caused problems when converting for PAL countries at 25fps. But this time, a broadcaster rejected the 25fps 50i conversions our digital lab was making

============

Question: is the footage you GOT ( 29.97p ) good for looking at ???  Is it OK when you look at it ??

Question 2: Is your goal to make PAL 25 fps ???

?????

If so, can you share an extremely small byte count file ( about 2 seconds ) using dropbox or google drive or something...so someone can check out what you got and what they have to do to convert to PAL ??

It seems to me that you should HAVE ( delivered to you by animators ? ) a video file that actually WORKS and looks good on your computer. It may be DNxHD or whatever, No matter... we don't care. It can be ANYTHING.  Can you share 2 seconds, despite the coprights etc. ??

[rodneyb56060189]

the reason I ask is.....

saying it's a telecine conversion from 23.97 to 59.94i ( which may very well be 29.97, but interlaced) would be FILM (plastic film) for NTSC broadcast. So that's pretty normal.  Going to PAL from 30 fps shouldn't be that difficult I don't think...but I'm no expert.

(getting another beer now )

[l93812963]

Answers...

Question 1: Yeah, the 29.97p video looks fine to look at. There is a little stuttering on the pan shots because of the hold frames introduced by the telecine/pulldown. It's not very apparent. But now that my eyes are so tuned to look for stuttering (with this conversion process), it looks bad to me now. But I didn't really notice it before.

Question 2: Yes, the goal is to make the 25fps. If you try to go straight from 29.97 to 25 with our footage, it looks extra jerky during pans. Because the shots still have the pulldown hold frames and it's also removing frames that might not be the hold frames during the conversion down to 25. And we also want to avoid converting it the other way where the frames are blending.

I wish I could share a clip. But our company is extra strict about this stuff.

Maybe I can shoot video and show what's going on in a Premiere project. I'll have to think about it.

I'm just trying to make this process faster than what we are doing now. We have 20 more episodes to convert in the next couple of weeks and eventually about 50 more after that. Maybe I need to have a long sit down with my editors to show them what I want so I don't have to keep fixing their work. But I'm thinking their eyes are just not tuned to this, even though it's just stepping through and looking for hold frames.

And to the reply above about using Terenex to detect cadence changes... I believe our digital lab was using this unit to make their conversions to no avail. They've told me they tried a bunch of different ways. Although, I wish we had one at work to try it ourselves. Seems like it should be able to fix this, at least according to the product web page.

Also another thing, we sent our 29.97p final master straight to the broadcaster's digital lab to see if they can convert it to 25p themselves and they couldn't ...or wouldn't that is, because it would be basically fixing every shot because of the of all the cadence changes.

[Trevor_Asq]

You stared with A B C D frames that were ‘telecined ‘ to A B bc cd D frames to get from 23.976 to 59.94i.

What made it easier to rotoscope (deinterlacing to 29.97P) makes it harder to reverse telecine (because you needed the fields from the bc and cd frames to remake the C frame.

Your ‘sped up’ result is presumably 23.976 being played at 29.97? If you interpret as 23.976 does that correct the speed?

If that hasn’t fixed it take a look at your 29.97P originals in Avid, having changed field motion to interlaced (bin column)

Assign step fields to keyboard shortcuts and step through field by field to see if the original fields are still there (ie 2 out of every 5 frames would show a difference between upper & lower fields - these would be the bc cd frames.

You can do same in Pr but harder to step fields.

If you do see temporal differences between fields of same frames in Avid then report back & I think there’s a slightly less cumbersome workflow (in Avid) to fix. It’s still pretty cumbersome mind

[l93812963]

How do I delete this thread? My co-worker found this and knew it was me. I might have violated company policy when I figuratively said I'd kill them.

[rodneyb56060189]

Some 'moderator' or special person might be able to do something, like lock the thread ( close it) or move it to another forum or something, but this stuff goes on the internet pretty much immediately ( so when you search for answers you can see adobe forum topics as possible answers ). I don't think anyone can make it like " it never happened" but who knows.

Offhand, I'd say that anyone who reads this thread can tell you are making a real effort to fix problems and make a workflow more productive and faster, for ALL INVOLVED.

I once told my little sister that I would kill her if she ratted me out to my mother, cause I shot off a firecracker in the back yard. It's not real, just an expression.  Everyone knows that. Don't worry. Be brave and honest and you'll be fine, in my opinion. Meantime maybe just ask that the thread be ended, so it doesn't show up at the top of the list of forum topics anymore.

Mark one of your own answers as "correct" and that will also keep people from trying to answer the question, and start to bury it in the forum archives.

Good luck. Don't worry.

[l93812963 · Accepted solution]

Thanks to the moderator for editing my post! Now if I get fired, at least it won't be for that.

So I'm thinking the convoluted process we are doing now in Premiere (render it 5 times -offset by 1 frame. Import into 23.976 sequence and pick the shot that is correct from the renders) is probably the best way to remove the pulldown, because the cadence changes so much. Sometimes even within the shot and during dissolves. Sometimes there were speed changes, mouth flap fixes that were made to different cadences. So it's all quite messed up. And most importantly, my editors and myself are way more familiar with Premiere than After Effects.

I've given up on trying to re-interlace the video to get Premiere to do whatever it's doing to be better at removing the pulldown. Tried so many things for a few days.

But mostly, I think the solution to my problem (going crazy) has been greatly eased. It seems like my editors really want to be on-board and want me to tell them what types of shots they are missing. Probably because they read this thread and know my frustration. So after talking with them and giving them tips to what to look for, their work has gotten way better. I can check them much faster now.

So I guess it was all on me for not really speaking up and communicating with the team. I should have been more clearer and not have been afraid to talk.

Thanks for everyone's input and advice.