Hunting ‘Jitter’ in Subnautica
I’ve been filming Subnautica for the purpose of creating Seamoth promo assets for several days now. For some reason, this video has been breaking my balls. I think I’ve discovered why, and suggest it may be something we should address.
Previous update videos have generally been much ‘slower’ than Seamoth. Habitat involved lots of static shots looking at static stuff like Moon Pools. Seamoth involves torpedoes, elecromagnetic defenses, and other fast paced ‘action’ stuff. I’ve found it really hard to get shots that ‘feel smooth,’ action sequences that flow, and shots that cleanly cut together. For example, I’ve got 100’s of GB of attempts at recording hitting a predator with a vortex torpedo.
Earlier today I got a particular shot which I thought was pretty cool. In in, I bop a Bone Shark in the face with a torpedo. I thought it might go well with a particular part of Simon‘s soundtrack as a ‘slo-mo’ shot. In creating the slo-mo effect, I stumbled upon something which I hypothesize explains why I’ve had so much trouble filming this update. Check out the slow mo edit:
httpv://www.youtube.com/watch?v=pZTq6_256MA
That edit is a 60fps video produced from 90fps source material. The slomo section plays at 20% speed, so 18 ‘real’ frames per second and 42 ‘fake’ ones, interpolated from the real ones. This technique reliably produces slick-ass slo-mo in other games, for example NS2.
In Subnautica’s case, it looks like festering garbage. The reason becomes apparent if you watch the same scene, frame-by frame, at very slow speed. Here it is at 2.5% speed, or 2.25 ‘real’ frames per second, with no ‘fake’ frames added:
httpv://www.youtube.com/watch?v=I2mGm5ot68k
Notice the way all the elements in the frame – The torpedo, the seamoth view model, the bone shark, the water effects, they are all updating at different rates. Each frame is a varying combination of updates from the various elements. At full speed, it’s hard for our eyes to perceive it consciously. My hypothesis is that our brains do perceive this mess subconsciously, and that Subnautica therefore does not ever ‘feel’ smooth.
This is not something that would be detected by our performance testing, because the game is updating at 90 fps, satisfying the performance benchmark.
If my hypothesis is correct, then I further argue that fixing this deep-seated issue with the game could improve our customer experience more than any number of new features. Though, I tremble in fear at the Pandora’s box of technical complexity this may open…