Matt's Homepage

There is a new stop-motion game like The Neverhood and Armikrog or recent movies from Laika. Harold Halibut is a adventure game that uses stop-motion and physically captured model objects to tell the story of a community that crashes their spaceship into a planet covered by the sea.

They made all the objects and stop-motion characters by hand, with real cloth, paint, sculpting, etc. When they realized how much work the stop-motion animation was going to be, the big idea was to 3D scan their hand crafted scenes, objects, and characters in the classic T pose, then use standard digital rigging systems to apply motion captured animations instead of painstakingly hand-animating every frame.

While this was a brilliant method to reduce the massive amounts of time and animation effort required, it still took them over 14 years to complete the game. They freely admit that most of that time was spent just figuring out the workflows since they weren’t well versed in game development tools. Still, what takes Laika hundreds of workers years was completed by this team with a fraction of that effort. They were able to add use all kinds of amazing effects and create scenes nearly impossible for true stop-motion animation.

Watching the resultant gameplay, some of the scenes are gorgeous. The close-ups and dialog shots are amazing and the facial animations are butter smooth. There are even tiny idle animations and movements that you would never do with stop-motion and a great depth to the game by letting you freely walk around – something impossible with hand-modeled animation. But there is maybe the only gripe: it’s too smooth.

Part of what makes stop-motion animation so quaint and ‘comfy’ is the little imperfections and limitations like clothing that interacts differently and animations that randomly pop and hitch. With this method, I notice the animations (especially walking animations) are a little too smooth and they often lose that stop-motion quality. There are times when they stretch the mesh too much and it becomes obvious the model is just getting stretched/bent. Individual clothing layers do not interact separately – they bend together as one. It feels like a solid plastic model – instead of having individually reacting layers of clothing/hair/etc. There is also none of the random occasional popping of clothing/animations mysteriously between 2 frames.

There could be ways to fix this by turning off random parts of motion blending between keyframes and having shaders that could randomly add some pop/hitching. Layers of materials could be animated separately. Still, it’s a noticeable distraction and difference between real stop-motion.

Also very noticeable is that the lighting is computed not physical. Especially in the larger/wider scenes, lighting is clearly rendered and it makes things look flat. Objects do not cast the physically correct kinds of shadows or receive mixes of soft and hard lighting edges as if the physical objects were place together and lit as a whole. This makes the rendered versions of the 3D objects (especially in wide shots) look flatter than they would if the real scene were physically created and lit.

This is definitely a novel new technique that is likely going to transform some of the industry. I think it has some amazing possibilities for speeding up dialog and closer-up shots; but probably not good at totally re-creating the aesthetics of stop-motion. I do think some of the smoothness/deformation and lighting issues could be fixed – but that will take a lot more work. Interestingly enough, Laika goes the OTHER direction. They computer generate/animate their faces in modeling tools, then physically 3D print them to put onto the objects into the physical world.

Articles:

• https://www.pcgamer.com/harold-halibut-is-a-handmade-stop-motion-aesthetic-adventure-about-friendship-and-home-in-a-giant-spaceship-trapped-beneath-an-alien-sea/