Testing dynamic frame rates in one 3D sequence

submit to reddit Oct 18, 2012 2 Comments by

Polish production company FlyFilm reveal the results of an investigation into whether increasing frame rates only for action sequences can deliver both judder free scenes and the ‘film-look’.

review dividing line Testing dynamic frame rates in one 3D sequence

img 413 poster 210x300 Testing dynamic frame rates in one 3D sequence

Established by DP/stereographer Piotr Kalkowski and writer/director Andrzej Stopa, FlyFilm were responsible for the world’s first HDR video advertisement and the first independent stereoscopic film in Poland which won Best 3D Film at the New York BeFilm Festival (pictured left)

After audiences reacted negatively to preview footage of Peter Jackson’s The Hobbit 3D, shot at 48fps, the company questioned whether the 24fps 'film look' and lack of judder (especially apparent during panning shots in 3D) were incompatible.

In an attempt to create a comfortable viewing experience and image that looks both smooth and cinematic, they shot 3D test footage at 50fps and then eliminated every other frame to create 25fps chunks of action in a 50fps file. The idea was to test whether one sequence can be filmed with various frame rates and compatible with a single projector. 

FlyFilm used a beamsplitter with two Canon 7D DSLR cameras for the test sequence which sees a jogger having her phone stolen by a man on skates.

As Canon 7D cameras do not have a 48fps option, FlyFilm used the 720p 50 fps setting for the shoot and then reduced the frame rate to 25fps in post-production. What is unique about the test, was that the frame rate was only reduced during scenes where the action was slow or where there was no camera panning. To achieve that, they dropped every second frame to get the 25fps and then duplicated every frame to get to 25fps in a 50fps domain. A simple fade masked the transitions between the standard and high frame rate footage.

The connection between exposure time and frame rate is the most important problem to consider.  The higher the frame rate is, the shorter the exposure can be to prevent that jittery Saving Private Ryan effect.

FlyFilm tells 3D Focus “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 convert 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 the 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. Let’s 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 it is 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 too sharp. This is exactly why Mr. Jackson and Mr. Lasnie used a 270 sector ( around 1/70s) in The Hobbit. They needed the movie to be watchable in 24fps so they had to compromise".

In third and final test on the video (below) FlyFilm 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 at a high frame rate and the difference when converting to standard frame rate isn't as jarring as in 1/200s.


The conclusion was that variable frame rates appears to work to deliver the 'film look' and reduce judder during action sequences, rather than apply a uniform frame rate to the entire piece.

In a statement, FlyFilm conclude: “The combined footage is in our opinion better looking than 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 dialogue 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 a high frame rate to properly register action parts. It means that slower parts need additional motion blurring applied.  We used motion estimation based on 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 the 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 colour gamut, but this does not mean that we have to show strong saturated pictures all the time. We can show black and white movies on a colour 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?"

Referring specifically to 3D, the results suggest: “It is safe to say that on average, 75% of a 3D movie doesn't need a higher frame rate; there are only some fast parts that require it. Choosing a frame rate for 3D is an artistic decision and variable frame rates gives a real artistic freedom in that regard. So there is no reason to approach this as an either/or case. 3D filmmakers should not artificially restrict themselves to a new standard that feels strange for the average viewer and costs a lot more in post, especially if the solution to this problem is relatively simple."

Piotr Kalkowski and Andrzej Stopa co-operated with Mr. Daniele Siragusano from CINEPOSTPRODUCTION

For more information about FlyFilm, check out their creative website.

What do you think? The following video is the left eye of the three tests. We would like to hear your comments below.




  • Cristiano

    could be interesting know how they get the genlock sync at 50fps with 2 Canon 7D DSLR cameras to do tes test…!

  • clydeD

    >>FlyFilm used a beamsplitter with two Canon 7D DSLR cameras for the test sequence which sees a jogger having her phone stolen by a man on skates.

    You lost me after that sentence. There was/is no point (beyond cursory curiosity) in reading about their experiments in HFR or DFR 3D, if the sensors are not in sync.

    I’ll give them credit for the first HDR film though.