Is Variable Frame Rate Better than High Frame Rate for 3D?
In the 24 frames-per-second vs. High Frame Rate (HFR) discussion, both sides have very good and well documented arguments. But Production company FlyFilm (Poland) thought that maybe there is a compromise to be made between that beloved 24 fps film look and lack of jittery motion that you can only get in higher frame rates?
To find out, they did a test and combined standard and high frame rate footage in one shot. They shoot 3D test footage in high frame rate (50fps in our case) and then get rid of every other frame to create 25fps chunks of action in a single 50fps file. You can see the results in the 50 fps video file downloadable here under (SteroscopicPlayer needed) and decide for yourself if FlyFilm managed to create a comfortable viewing experience and image that looks both smooth and cinematic. The test is extensively described here under by its authors, Piotr Kalkowski and Andrzej Stopa from FlyFilm.
Test Sequence Setup
We used a very basic setup, just two DSLR's in a beam-splitter mirror rig. Canon 7D's cameras do not have 48 fps option, so we couldn't re-create the 48/24 difference that everybody will be able to check out for themselves in upcoming Peter Jackson's "The Hobbit". Instead we used 720p 50 fps setting for the shoot and consequently reduced the frame rate to 25 in postproduction, but only in the parts where the action was slow, or the there was no camera panning. To achieve that, we dropped every 2nd frame to get the 25fps and then again duplicate every frame to come to a 25fps footage in a 50fps domain. Then we edited the material in 50fps timeline. For transitions between standard and high frame rate we used a simple fade. For the presentation we choose 1/100s exposure time, and that was the most difficult decision.
The connection between exposure time and frame rate is the most important problem to consider. Basically the higher the frame rate is, the shorter the exposure can be without experiencing that jittery Saving Private Ryan effect. You can go with 1/200 at 50fps as we did in the first test and have great, clear frames, but it will also mean that if you decide to converse it to 25fps it will look very jittery, just like in the second example. There will be very noticeable difference if you try to combine those two in one movie clip. On other hand there isn't much point in using standard 1/50s in high frame rates, because the difference isn't worth the extra frames you have to render. You will still get that annoying motion blur. Lets convert exposure time values in those examples to degrees of sector. As we know, 180° is normal, meaning for 50fps it is 1/100 and for 25 fps 1/50. So 180° is desired. But shooting in 50fps with 180° (1/100) and then converting to 25fps will lead to a sector of 90° which is to sharp. This is exactly why Mr. Jackson and Mr. Lasnie used 270 sector ( around 1/70s) in Hobbit. They needed the movie to be watchable in 24fps so they had to compromise. In third and final test we decided to use 1/100s exposure time (180° in 50fps and 90° in 25fps), because it takes good care of motion blur, looks quite good in high frame rate and the difference when converting to standard frame rate isn't as jarring as in 1/200s.
Download the 3D Test Movie
The test movie is stored as separate left and right mp4 files (VFRtest_L.mp4 and VFRtest_R.mp4, 137 Mbytes each file) in an archive you can download from sendspace here. You will need a 3D movie player such as StereoscopicPlayer from 3DTV.AT to play it. A monoscopic 25 fps version is on the 3dfocus channel on YouTube to give you an idea of the content (but it is of course impossible to check the 25/50 fps differences there!).
The basic idea of variable frame rates appears to be working as far as it brings out the best of two worlds. The combine footage is in our opinion better looking then either exclusively 50fps or exclusively 25fps footage. The simple fade technique also worked quite well but, as previously stated, the exposure time is a problem. A knowledgeable viewer will see the difference and jitter that appears in 25fps parts.
The simple solution is to shoot the parts you want to show in 25fps in that frame rate with appropriate shutter. Later in post you just double up the frames and end up having a 25fps in a 50fps time domain. Most conversation scenes and static shots could be done this way.
But shots that combine action and static, like in our example, have to be shot in high frame rate to properly register action parts. It means, that in slower parts applying additional motion blur is necessary. We used motion estimation based motion blur algorithms. We turned the 25fps shutter back to 180° in post. It takes a bit more rendering but it's working very well and it is hard to notice jumps between frame rates.
In our opinion, we mustn't force every shot to use the whole temporal resolution. A projector might have a big color gamut, but this does not mean that we have to show strong saturated pictures all the time; we can show black and white movie on a color projector.
Why not show a standard frame rate for some shots or even for some parts of one shot on a high frame rate projector? It is safe to say that 75% of average 3D movie doesn't need higher frame rate; only some fast parts require it. Choosing frame rate in 3D is an artistic decision and variable frame rates give a real artistic freedom in that regard. So there is no reason to approach this as either/or case.
Stereoscopic 3D filmmakers should not artificially restrict themselves to new standard that feels strange for average viewer and costs a lot more in post, especially if the solution to this problem is relatively simple.
FlyFilm is a production company offering stereoscopic assistance at every stage of 3D production. Established by DP/stereographer Piotr Kalkowski and writer/director Andrzej Stopa, FlyFilm places great emphasis on testing and implementing new technologies. Since early 2009, they shot the world's first HDR video advertising, one of the first web series and interactive advert.
They also made the first independent stereoscopic film in Poland, 4:13 to Katowice (Watch the 3D trailer, also on our 3D Movies Trailers page). The stereoscopic thriller won for BEST 3D FILM at New York BeFilm Festival.
Thanks to Cinepostproduction
This paper was written by Piotr Kalkowski and Andrzej Stopa and published with their authorization, but they would like to thank Mr. Daniele Siragusano from CINEPOSTPRODUCTION (Germany) for its enormous help in finding some of the postproduction solutions.