Digitize Super 8 film to 18fps or 25/30fps?

The service did the test clips for me.

Neither Apple ProRes nor the usual applications (FCP, Premiere, After Effects, DaVinci Resolve) do not seem to support 18fps (*). So the service exported each scanned frame to a 25fps 1920x1080 ProRes 422 HQ file (actual image area is 1444x1080 with 238 pixel black borders on both sides). That ”raw” ProRes material is further edited so the 139% fast-motion (18fps played at 25fps) frame rate was not important at this step.

I can edit and slow down the clips in Final Cut Pro X to their normal speed. For such a (Super 8) 25fps project, 72% ”slow-motion” yields the normal speed (18/25 = 0.72). That same 72% applies also when that 25fps ProRes clip is dropped to a 30, 50, or 60fps project while a 24fps project needs 75% to keep the project length exactly the same with normal speed. In each such project just the correct number of frames are added with the desired method (duplicate, blend, optical flow). No extra generation loss.

The service routinely makes also a 25fps H.264 25Mbps MP4 delivery version with 18->25fps conversion with a chapter every 3 minutes with the correct speed and duplicated frames (these are silent films but they used their usual 71.85% “slow-motion” that fits Super 8 audio to the total project length so the actual frame rate is then about 17.9625fps).

I now recalled that one old Super 8 project was actually shot at 12fps for a comic effect. If needed, I can now easily adjust its speed to normal with minimal extra fps conversion artifacts with a 48% slow-motion in FCP.

The quality was good (the low-res demo clip link in my earlier post is old and was not done with their current Rank Cintel 4K hardware).

(*) With great effort and lots of trial and error I managed to make an animated 60 second 18fps MP4 test pattern clip via Photoshop, FCP, QuickTime Player 7's 24fps Image Sequence with a pattern of duplicated frames, that were then stripped off via VLC's conversion to 18fps. My LG OLED TV played it OK.

Based on the service's low-res demo clips (link below) they convert 18->25fps by duplicating some frames with some pattern so the movements don't accelerate. They don't seem to blend frames.

Final Cut seems to do basically the same if a 18fps clip is dropped to a 25/30fps project, right? If correct, then I could have best of both worlds by having "raw" ProRes 422HQ at 18fps and let Final Cut convert it to 24/25/30/whatever frame rate I need in the future. If I now let them convert 18->25fps, then a possible further 25->30fps conversion will produce even more jerky movements. I am not sure how long PAL countries continue to use 25fps with the push from mobile devices defaulting heavily at 30fps.

Final Cut's "Automatic Speed" seems to convert frame rates by lengthening or shortening the clip but not duplicating or skipping any frame. So 100 original frames at 25fps is 4:00 (4 seconds) and the same original 100 frames with "Automatic Speed" is 3:10 (3 seconds and 10 frames) at 30fps.

http://www.kaitafilmi.fi/super.htm

The service did try to output 18fps but they said the exported file didn't seem to have that frame rate so they settled for a more proven 25fps for that "raw" frame-by-frame ProRes 422 HQ file.

I made 1 minute test clips from that Super 8 18fps frame-by-frame digitized footage by converting it with Final Cut Pro X 10.4.8 (*) to .m4v:

24fps (repeating 112 duplicate frame pattern)

30fps (repeating 122 duplicate frame pattern)

25fps (repeating 12 112 12 112 112 12 112 duplicate frame pattern)

I played the clips on a 55" LG OLED TV and on a computer monitor via QT Player 7 (the TV has some motion adjustment settings so I tested this also on a computer monitor).

I can often spot uneven movements caused by duplicated or skipped frames in 25<->30fps conversions from high quality DSLR or iPhone footage. I was surprised I couldn't see practically any difference in these Super 8 test clips. Maybe my hand-held low-quality Super 8 test footage was more forgiving.

(*) the frame-by-frame source is a 25fps fast-motion ProRes 422 HQ file. FCP sees its speed in a 25fps project as 100% so it must be slowed down by 72% (18/25 = 0.72) for normal speed.

When that 25fps source file is dropped to a 30fps project, FCP keeps the duration and 100% speed the same by duplicating frames behind the scenes so it must be slowed down the same 72% for normal speed.

24fps behaves slightly differently because FCP automatically slows down 25fps clips by 96% when they are dropped to a 24fps project (i.e. no skipped frames). So such clip must be slowed down 75% (0.96 x 0.75 = 0.72) for normal speed.

...The service routinely uses 71.85% slow down for such Super 8 25fps conversion (i.e. 17.9625fps) to keep the possible audio in sync and I couldn't see any difference in it either (such clips have a slightly different duplicate frame pattern that might not repeat at all).

Those fps conversion Terminal commands seem to work properly on video-only files -- an audio track seems to make the video stutter.

AFAIK my service tried to use Premiere to output 18fps but the exported frame rate was incorrect. The links below indicate that although it is possible to make a 18fps Premiere project and export 18fps, the exported file might be forced to 15fps especially with the H.264 codec.

It seems that currently only applications like DaVinci Resolve, After Effects and Scratch support old frame rates like 18fps.

One comment in the links below claimed that "clean up" filters are much less effective when there are duplicate or blended frames. So even if the service did output 18fps ProRes 422HQ, my FCP app would then automatically convert it to a 25/30fps project with duplicated frames. So I'd then have to apply "automatic speed" to get unique frames, apply filters etc, and then revert back to normal speed with duplicated frames, right?

On the other hand, if I have fast-motion unique-frame ProRes 25fps, I could edit the material, and as a last step, apply 72% slow-motion for 25/30fps projects and 75% for a 24fps project.

Another option would be to switch to and learn DaVinci Resolve and use the smooth original 18fps also as the final mp4 delivery format because the computer monitor and my TV accept it. And convert to 25/30fps only if needed. But it would be a steep learning curve with little real benefit.

p.s. in the last link there is an interesting video about mixing Super 8 to separately recorded audio: "Since Super 8 cameras don't run at real-time unless they are crystal-synced, you have to splice up and squash and stretch segments of the video (NEVER the audio since that's realtime) in order to get it to match up to the audio".

https://community.adobe.com/t5/premiere-pro/export-18fps-sequence/td-p/10486209?page=1

https://forum.videohelp.com/threads/374344-Can-Premiere-Pro-CS6-export-at-18fps

https://cinematography.com/index.php?/topic/77756-18fps-in-a-25fps-timeline/

Thank you all for your insightful comments.

I might first let the service digitize a small sample at 18 and 25fps, do some tests, and then decide which one to use for the remaining 4,5 hours of film (just old home movies).

I made a 1 minute test movie at 15fps from the sample test pattern movie files at the link below (I made 1 minute 30fps movie and exported it at 15fps from MPEG Streamclip. I couldn't find a workflow to produce 18fps -- even QT Player 7 Pro doesn't have that image sequence option).

LG OLED TV played my 15-25-30-60fps test clips with no problems so I hope it can handle 18fps, too. If not, I might use Final Cut to do the frame rate conversion.

Just for fun I did generations of 25-30-25-30-25fps conversions and that last back-to-25fps generation was awful with many back-to-back duplicated and skipped frames. I would rather have no generation loss from unnecessary frame rate conversions so 18fps as a master file might be the best starting point.

About "Automatic speed": In Final Cut Pro 10.4.8 above the timeline there is the clock icon with Choose clip retiming options > Automatic speed. If you drop a 60 second 25fps clip into a 30fps project it stays 60 seconds with those occasional duplicated frames. Choosing Automatic speed shortens the project to 50 seconds with no duplicated frames. Choosing Custom... > 120% speed or 50 second duration has the same effect but needs some calculations from the user.

There are some Test Patterns for different frame rates (no 18fps though) at the link below:

http://forum.cyberlink.com/forum/posts/list/25402.page

FWIW, DaVinci Resolve v14.2 added support for 16 and 18fps. I just tried v16.1.2 and it seems to output 18fps ProRes 422 HQ OK (I might still let the service output to 25fps fast-motion unique frames ProRes if that is their proven workflow).

I could also successfully losslessly convert 25fps H.264 to 18fps with this tip:

Change frame rate without re-encoding. It requires remuxing the file (mp4, m4v) to a different containter format MKV and then remuxing it back into an mp4. Here is an example that changes a video to 18fps:

mkvmerge --default-duration 0:18fps --fix-bitstream-timing-information 0 original-video.mp4 -o temp-video.mkv

ffmpeg -i temp-video.mkv -c:v copy slow-video.mp4

...install instructions in macOS 10.15:

/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"

brew install ffmpeg

brew install mkvtoolnix

https://superuser.com/questions/320045/change-the-frame-rate-of-an-mp4-video-with-ffmpeg

http://forum.blackmagicdesign.com/viewtopic.php?t=57059