A machine learning model helped achieve the effect.
I’m looking forward to hopefully bringing you some in-depth coverage of Dune: Part Two in the coming weeks and months. It’s an incredible film.
In the meantime, here’s an interesting piece of info from production visual effects supervisor Paul Lambert about turning case eyes blue. This time around, the VFX team utilized a machine learning model to help them craft the blue eyes.
This quote from Lambert is from the official production notes for the film.
“We have a multitude of different techniques we used in this film, one of those being that because we had a lot more Fremen characters, we have a lot of eyes to make into that beautiful blue, and for much more of the film than in the first one, over a thousand shots. We came up with a different technique, using what we’d learned before from the hundreds of blue eye shots in the first movie and creating a machine learning model, an algorithm trained from those ‘Dune’ shots to find human eyes in an image, which would then give us a matte for the different parts of the eye. We then used this multi-part matte to tint the eyes blue. Some worked better than others, those others we did by hand. It actually went full circle sometimes, in that we had to take out some of the blue eyes that got generated in the non-Freman characters, rather than add them, as the algorithm would just find eyes, whether they were Freman, Harkonnen or Sardaukar! But, it was a brand new technique, getting that done.”
