This article is also available in:
French
3DVF: Water simulations are getting more realistic year after year, and the tools we use get better and better. Which improvements are you hoping for in the near future?
Florian Gellinger – RISE: It would be nice if GPUs had significantly more memory so that you could do more of that work interactively, just like some of the fire and smoke work is done these days. And if that FLIP solver would then recreate the same but more detailed results on the CPU for final cashing – everyone would be in love.
3DVF : How did you create & animate the eels in the aquarium?
We started with a rigged eel that we hooked up to our Houdini crowd simulation to fill the aquarium with hundreds of eels. It quickly became too crowded and you ended up with Hugh Jackman’s face completely obscured in the tank. So we reduced the number significantly and then had to hand-animate roughly 50% of the shots to have more control over timing and framing in close-ups. Compositing supervisor Oliver Hohn and his team balanced all the shots and made sure lighting inside the tank matched the face of Hugh Jackman – but grading the shots neutral and matching skintones was difficult because Hugh Jackman’s face under water changed color depending on how little air he had left 🙂
3DVF: Can you also tell us about the stunt doubles?
There is a shot where Hugh Jackman jumps from a roof, chasing the villain through a sunken recycling plant, dropping down one floor and landing hard on a steel railing on the opposite side of the canal.
The shot starts directly in Hugh Jackman’s face and as the camera dollys behind him, we switch to a 200% screen height digi double who takes the leap for him.
In order to build the transition properly, we sculpted the wrinkles of the CG shirt to accurately match the ones in the plate before FX takes over with a cloth simulation.
The environment is full CG, picking up our surroundings from earlier shots but giving the team for freedom of movement with the camera by not having to shoot this on a rooftop.
3DVF: Due to the pandemic, the team had to work from home. How did you approach the communication within the team?
We started to use Zoom throughout the day, opened individual Zoom rooms per department or sequence. This enabled us to feel almost as if we were in the same room. However, this way of working is terrible and it does not educate less experience artists by learning from others. In a normal year I would take a walk to grab a coffee, looking at everyone’s screen and getting a feeling for our progress on the show. I could see if artists are doing well or if they need help. I could see if artists prioritize correctly and I was picking up that sort of info effortlessly. VFX is a team sport and the more info the team picks up during the day, the better for their game.
3DVF: One last question. New technologies are emerging: deepfakes, neural rendering… Is this an area RISE is currently exploring?
We hate all new things and oppose anything that looks innovative – just kidding. We’re currently making a big move towards 100% USD workflow, have recently switched to Renderman, swapped out the servers and are introducing a lot of new tools and workflows necessary to manage and maintain 5 facilities. That’s already a lot to deal with. We’ll look into other areas when our clients demand it. We’ve done some realtime camera tracking and Unreal Engine work for a big production recently – and we’d do it again if there is demand. The tools are all there and we’ll make it work if the content requires it.
More Information
