Some time ago, I've started modeling a high-poly Stargate for a project of mine in 3ds Max. The idea was to create a 100% accurate model of the screen used prop from the TV show
Stargate SG-1 (I'm that big of a fan). As I already expected, over time it became a bit more difficult to maintain accuracy as I got into modeling the more obscure parts of the 'gate. For those parts, I just eyeballed things or extrapolated from the references I have.
In its current state, it's not yet finished (at a whopping 2,519,603 tris and 1,323,806 vertices). There are a few minor details to complete, but I decided to do the low-poly version first, and those details can easily be painted on. I'm not sure whether the model is suitable for normal baking, but I've already got a plan B and C if things go wrong. Lighting for the chevrons will be baked using V-Ray (with which these renders were also made).
I'm not yet certain how detailed I will make the low-poly version. I'm thinking about using low-poly glyphs on the inner track, rather than using a normal map, so close-up and profile views look good. And since it's the main focus of most environments in which it stands, I think I can get away with a few more tris (the current model has a little over 9,000 tris without the glyphs, divided into the base, the glyph track and one chevron which is reused 9 times). I'll of course LOD the hell out of it for optimization later on.
I plan on using Substance (Painter & Designer) to texture the model, then bring it into Unreal Engine 4 and start coding things in C++ and Blueprint to make it work (at least partially).
Eventually, I may chronicle the process on creating the Stargate on my blog, also for me to reflect upon on what I learned.
Replies
You got some pretty small details modelled.. yet some of the bigger shapes still look pretty sharp.
Those edges will have a bigger impact on the overall look than the super tiny details being exact.
I get what you're saying, but I don't go around subdividing everything if it isn't necessary, and at this point (as it's not finished) I haven't beveled the edges yet. But even then, on the real prop, the round edges are barely noticeable; they're there of course (nothing has real sharp edges, I know that much) but they're barely noticeable compared to the other details, so currently they're not that important to make.
The 2 million tris come from all the embedded details. The rectangles with the diagonal curves were subdivided to smooth them out, just as the other curved details and then removed unnecessary polygons so it's easier to extend the model outward. The glyphs were made in Illustrator, then imported as splines (with Interpolation set to Adaptive). I used the Bevel modifier to extrude them.
Here's a quick render showing the wireframe. There are numerous N-gons, but they don't cause trouble as I've checked their triangulation, and they're all mostly flat.
It's a bit over-the-top, but the glyph track is LOD 0. I plan on having it switch to a normal mapped model at about 5 meters (which would make it 156 tris). It's there for when you're right in front of it.
I may be able to reduce it some more, but I don't think I can make it less than 10,000 tris for LOD 0. I'm still not entirely sure what a proper workflow would be for LODs (multiple textures/materials for example).