If you want to know how much work it takes to create a big budget game, here’s a quick indication of the work involved.
The studio behind the excellent Spider-Man, Insomniac Games, gave a post-mortem at the Game Developers Conference this year. It was similar to what many studios do after a major release, breaking down the challenges they experienced, how Spider-Man‘s procedural generation worked and what precisely was generated, finer detail about the game engine, and everything in between.
There’s a lot of incredibly dense technical detail. But there’s also bits and pieces that are easier to parse, particularly this little nugget towards the start that outlines just how many assets make up Spider-Man‘s world.
The talk also put a figure on the extra level of detail that went into the models, as well. Using Morgan Michaels as an example, Insomniac senior engine programmer Elan Ruskin explained that Michaels’ gameplay asset had just under 5200 triangles, while the cinematic asset – and one of the major reasons why Spider-Man‘s faces were so exquisite – had almost 90,000 triangles.
[referenced url=”https://www.kotaku.com.au/2019/01/some-appreciation-for-the-faces-of-spider-man/” thumb=”https://www.kotaku.com.au/wp-content/uploads/sites/3/2019/01/spiderman-2-410×231.jpg” title=”Some Appreciation For The Faces Of Spider-Man” excerpt=”Over the holiday break, I took pains to finish my extended playthrough of Insomniac’s excellent Spider-Man. It ended up taking me substantially longer than anticipated, but it did afford me the opportunity to enjoy the game’s superb range of faces.”]
It hits topics like the facial animation system implemented for Spider-Man – the skeletal animation for Ratchet & Clank was a deliberate part of the aesthetic, but they needed a system that would communicate the nuance and finer detail that Spider-Man needed. The bandwidth required just to stream the animation for faces in cutscenes was extremely high, but Insomniac were able to cut all of that out with some clever observations about what is and isn’t required during a cutscene – skin doesn’t move a great deal, so they were able to save bandwidth and compute time by calculating the normal and tangent vectors for faces in cutscenes in real-time, as opposed to just streaming it in (which would normally be done to save load on the CPU).
And naturally, there’s always something that goes wrong:
You’ll get the most out of the talk if you have some programming experience, and slowing the video down can help (Ruskin talks rather fast). But there’s also a sequence of videos after 49 minutes where things go hilariously, horribly wrong, like a horde of pedestrians running face first into a car, exploding cars launching into the air and then hovering around, and a Brooklyn bridge that cars just … do a U-turn on at the end.
Leave a Reply