Pulldown pattern for transferring 18fps to 25fps?

Patrick Cooper · February 19, 2016

Ive been googling for a while about this query but I just cannot get a definitive answer. I am enquiring about transferring super 8 18fps film footage to 25fps progressive video. I'm probably going to end up with a 1280 x 720p (25fps) video. I'm not sure what pulldown pattern would produce the best results because peoples' opinions on internet forums vary. Even Wikipedia wasn't much help. One individual says to duplicate every third frame twice. And someone else recommends that the first frame is displayed four times, the second frame is displayed twice, and the third frame is displayed twice - essentially 4:2:2. Just wondering which would give the best result in terms of smooth, natural looking motion? I know you'll never get a perfect result going from 18fps to 25fps so I guess I should ask which option gives the most acceptable results. Also, would anyone recommend any other pulldown patterns for such frame rates?

Roberto Pirodda · February 19, 2016

when i put the 18 fps footage in the 25fps timeline of FCP , it automatically place a 2:3:2:3 pulldown, and it is the best smooth looking to me. But if you want to improve the smootness you can do this :

place your footage in the timeline base layer, then place the same footage in the upper layer but shifted of one frame and with the opacity set to 50%. In short you will get the same footage mixed with itself, a sort of frame blending. You will notice moreover that the grain has less appearence, and there is less flicker ( if any).

Phil Rhodes · February 19, 2016

You can also use any of various optical flow interpolators, which will do their best to make up intermediate frames.

Freya Black · February 20, 2016

It depends on what you are going to do with it.

For myself I might be tempted to try one of the optical flow thingys that Phil mentions. Maybe Twixtor or that Warp thingy in AE or something. Isn't there something in FFMpeg or something of that nature that does this too?

Alternatively I might try and double up the frames to get 36fps and then slow it down to 25fps slow-mo.

Or maybe combine the two by doubling the frames in Twixtor or whatever and then slowing it down to 25fps.

Of course you might not want a nice smooth slow-mo effect.

This is why I much prefer 12fps!

Freya

Perry Paolantonio · February 20, 2016

In a word, don't. Have it transferred at 18fps, and then for the material you want to edit, pull up to your target frame rate.

If the question is, as it seems, about how to transfer directly to 25fps, I wouldn't recommend that. We're working on a massive collection of 16mm home movies right now, shot at 16fps over 35 years. It's all being transferred to 16fps files. Once that's done, there's a 1:1 mapping of the film to a digital file, with no baked-in interpolation to worry about. For viewing on a computer, this is ideal. If the client wants to make a DVD or Blu-ray, he understands that the files will need to be pulled up to a standard broadcast frame rate to do so.

Also - why a 16:9 aspect ratio for Super 8 (unless it's Max8)? At minimum, I'd recommend 2k, because that gives you a scan that matches the aspect ratio of the film, and is large enough that if you wanted to crop a 16:9 image out of it, you have a bit of compositional wiggle room.

-perry

Freya Black · February 20, 2016

In a word, don't. Have it transferred at 18fps, and then for the material you want to edit, pull up to your target frame rate.

If the question is, as it seems, about how to transfer directly to 25fps, I wouldn't recommend that.

It's common for people to get 18fps footage transfered at 25 or even 30fps, so the footage runs fast but doesn't have generated frames. Then in post they can interpret the footage as 18fps. There is still the issue of how to make a DVD etc however once you have your footage at 18fps. With internet streaming you can keep the footage at 18fps I guess?

Perry Paolantonio · February 20, 2016

Yeah, as long as the frame mapping is 1:1, the end result would be the same, but with an extra step involved, assuming that ultimately you want it to be the correct speed within a 25fps container.

Patrick Cooper · February 21, 2016

Thankyou all. Lots of options here. I don't really want to have the footage sped up or slowed down. Like I mentioned in my opening post, I would prefer to replicate natural looking motion more or less.

Also - why a 16:9 aspect ratio for Super 8 (unless it's Max8)? At minimum, I'd recommend 2k, because that gives you a scan that matches the aspect ratio of the film, and is large enough that if you wanted to crop a 16:9 image out of it, you have a bit of compositional wiggle room.

-perry

Why have the finished video in a 16:9 aspect ratio? So that I can play it back on a HDTV. But there's no way that I would crop the image to 16:9. What I want is to retain the whole image and have black pillars on either side, hence creating a 16:9 aspect ratio. So the original aspect ratio of the super 8 frame will be preserved (which is probably about 4:3 or thereabouts) and have that inside a 16:9 video file - if that makes any sense?

Perry Paolantonio · February 21, 2016

Why have the finished video in a 16:9 aspect ratio? So that I can play it back on a HDTV. But there's no way that I would crop the image to 16:9. What I want is to retain the whole image and have black pillars on either side, hence creating a 16:9 aspect ratio. So the original aspect ratio of the super 8 frame will be preserved (which is probably about 4:3 or thereabouts) and have that inside a 16:9 video file - if that makes any sense?

If you transfer the film to 720p, your actual image area in the scan is 960x720, more or less. If you were to think of the scan in terms of megapixels, that's about .6MP -- if you were to scan to 1080p, your image area would be about 1.5MP. If you were to scan to 2k, which is only slightly larger than the 1080p HD size (204xx1556 vs 1920x1080), you're getting more than double the resolution of a 1080p scan, and about 5x the resolution of a 720p scan. Even if you only ever want it at 720p, you're going to get a better image by scanning larger (2k, for example) and then scaling down.

Also, when you scan a 4:3 image to a 16:9 format, the operator of the scanner has to make the call about how it's cropped. Let's say you have a hair in the gate, or you shot on multiple formats and the gate position varied reel to reel - the operator of the scanner would need to zoom in a bit more than you may want, in order to make sure you get clean edges. By scanning at 2k, and doing a full aperture or even slightly overscanned image, you put that creative control back in your hands, and can reposition or crop as you see fit. It's a minor detail, but it's something that comes up a lot, especially with footage shot on multiple cameras where the gates will often vary in position a bit.

Lastly, if you ever wanted a 1080p or higher version, you'd need to scale your 720p transfer up, which is never desirable. Always better to scale down, if you have to scale at all.

Patrick Cooper · February 21, 2016

Even if you only ever want it at 720p, you're going to get a better image by scanning larger (2k, for example) and then scaling down.

If everything comes together right, I'm going to do the transfer myself. And I won't be scanning at 720p. I'll be using a Micro 4/3 digital stills camera to take a 4000 x 2248 photograph of every single film frame. The imported digital photo files will likely be Raw and will be processed in Adobe Lightroom. They will then be assembled together in VirtualDub to create a high resolution video file. At the last step of the process, the video file will be converted to 1280 x 720p.

Edited February 21, 2016 by Patrick Cooper

Dirk DeJonghe · February 21, 2016

In a Quicktime container the projector speed is just an number in metadata. If you capture at 24fps and change metadata projector speed to 18fps, it will playback slower but all frames will still be 1:1. It is only when you import the file into your timeline that fps have to match if you want to maintain 1:1. Obviously a clip running one minute at 18fps will be much shorter in duration at 24 fps while keeping the same number of individual frames.

Carl Looper · February 26, 2016

Yes, Dirk is correct.

The ideal solution would be to ensure the deliverable media file simply has a playback rate set at 18fps, where each source frame is represented just once in the target media file, ie. no duplicate frames and no motion interpolated frames. The playback rate for a media file is, as Dirk mentioned, a single number in the meta data. The media player just plays the frames at the rate specified in the metadata.

I was working on a project where material had been shot at 200 fps, with the idea of playing that back at a slower rate for a slow motion effect, but the developer couldn't get it to play back at a slower rate. It was playing back at 200 fps because 200 was in the metadata. He was trying to re-render the entire file in an edit suite for the desired effect. The far simpler solution was just to change the frame rate in the metadata from 200 to say: 25, to obtain the required effect. Which is what I did.

But exceptions to this are really quite common. The target media file might be required to play at some rate other than 18fps. Reasons for this might be limitations in the media format (eg. it doesn't support 18fps), or one might have a project in which other material, shot at a different rate is to be equally accommodated. Duplicating (or indeed skipping frames) according to a rate conversion algorithm, or creating motion interpolated frames (from optical flow), etc. become the solution here.

C

Edited February 26, 2016 by Carl Looper