Somewhat ashamedly Dave sent this instalment through at the end of last week, but due to me (uncharacteristically) being away from my keyboard I have only just got around to posting it now! This third section will discuss a topic which Dave and myself have spent a great deal of time talking over – workflows. Passing over to Dave now to elucidate a little more. As always feedback, comments and discussion is encouraged!
I’ve been wracking my brain over the past few weeks on how to present this next part of the “You Cannot Be Accurate” Series). As the title says, it’s the workflow that an artist/group of artists needs to do to get from an object that you would find in a museum to getting that object into the game.
I’ve thought about going into detail on each stage which then in turn would mean me talking about PBR shaders, the difference between realtime and pre-rendered and in my head, this series would end up growing and growing to basically cover game development. It would become unwieldly to not only myself but to you, the reader.
Ultimately I want to get the message across that the 100% accurate issue is a flawed notion and in this series, I hope I’ve so far explained bit by bit as to why it is. So I thought I would keep this post relatively simple by outlining in broad strokes the workflow and what I, as an artist, would have to go through.
Each stage has many issues and pitfalls which ultimately chip away at the fabled 100% accurate figure. So for example, sometimes photogrammetry/scan data can come out a bit lumpy/noisy in parts, or maybe there’s holes in the mesh which you can’t physically get to in real life. So you clean it up, by eye, as best as you can….a guesstimate. See? You’ve lost (if an amazing artist does it) at least 1% of accuracy by doing this.
Another issue is the nature of game engines and games themselves. They require lots of smoke and mirrors, and optimized, low polygon meshes. If the original skull scan I did was say 4million polygons then the final skull mesh which is now in game is 300 polygons at best. The normal maps, specular maps do their best to mimic the detail from the original high polygon model. but come on, 100% accurate…really?
And it’s the little things like this along this workflow that will lessen the “accuracy” of the model…it might be tiny, a pixel or polygon off but sorry, no matter how you look at it… it’s no longer 100% accurate anymore.
Now there’s far better people than me that can scan, clean up data etc (I’m a jack of all trades) but does the skull I made in the workflow look like the original skull? Is it 100% accurate?…of course not. Maybe if you want to get picky it might rate as 80% accurate at best…but does it LOOK like the skull? Isn’t that….kinda okay?
Are the 3d scan data that museums have paid thousands for, (currently sitting in servers doing nothing)…. 100% accurate?
So maybe we need to change out terminology a bit…instead of saying “I want this model to be 100% accurate in game”, maybe we should be saying “I want this model in game to be a good recreation of the original object”? Just a thought.
Time for you to have your say! Do you think accuracy of recording (millions of polygons and points) is a neccesary goal? Or is it more important to think contextually – to record a good aproximation of the context and item with supporting data? Vote in the poll and leave a comment below!