[WIP] Kirkbride Insane Asylum 3d Level Design for Mystery Game in Blender.

[Robbie]

I'm picking up where I left off in my WIP mapping thread since the 3d modeling has begun. Figured I'd post model work here and map work there. Here's an update to the modeling. This is the work ward on the male side with windows and doors properly cut out/in. It was another great learning experience, and I'm really picking up this blender thing. Unfortunately, the quickest way i could discover to do the windows was to use a boolean operation, which means I'll have my work cut out for me when I work on the full wards and admin levels and such. Boolean requires quite a bit of clean up after the fact, but nothing I can't handle.

3dWIP_02.jpg

So, even though this is a small section and only used in one other place (the female side) it was a good test, which I'll probably keep. I'll do an actual ward next since I have a process. That part can be re-used 16 times (8 normal, 8 mirrored).

Edit: Bonus test image of work ward duplicated to three stories. This ward will only be on the ground story though...there will be another slightly different layout of this on 2 and 3 but it shows the scale of the ward with the windows and doors in. No ceilings or floors yet.

3dWIP_03.jpg

[Mouse]

That's looking great Robbie

I don't know anything about how game engines work at all. What level of detail do you have to get down to with the 3D model - can a lot of the detail be created with the textures and normal maps, or does the model have to be complicated enough to show door handles etc?

[Robbie]

I'll be modeling a detailed door and window and lots of detailed furniture objects and props but they'll all be instanced and placed in the model in the game engine. Other than visualization purposes there's really no need to duplicate the door and window models in blender. Model once and reuse. I can use shaders in the game engine to make randomized wear and tear on instanced geometry. Pretty cool I think.

Sent from my HTC6535LVW using Tapatalk

[Mouse]

The more you say, the more it fires my imagination, and the more eager I become to see the finished game

[Robbie]

Me too! The idea I think is going to be amazing if I can pull it off. Thanks for saying that!

Sent from my HTC6535LVW using Tapatalk

[Redrobes]

I dont know which game engine your targetting or in fact much specific knowledge about game engines but I wondered about something which may be relevant here. I did work on 3D video drivers for a while tho so it may be pertinent. In most video engine optimizations its possible to work out that some polygons cannot be visible quite quickly without having to go through the per pixel depth cull. So is it necessary to break up long walls into shorter sections ? I wondered this because if its possible that a bit of a large polygon may be visible then it does not get culled quickly and the whole thing may need to be rasterized and then each pixel tested for visibility in the shader. If you break up a long wall or a long ceiling into parts then maybe most room walls and ceilings get culled leaving only the ones which may possibily be visible. Is that something that may help ? I wonder this at this point in case it means that your model could be done differently. I dont know the answer to this and I suspect you would need to ask someone who has a lot of experience with them. But I guess you could test if the engine renders it faster to find out.

I do know quite a bit about boolen cleanup since I wrote an implementation of an algorithm to do that some time ago. Writing 3D boolean is a lot tougher than it appears in case anyone is wondering. Tough to the point of you just want to brain yourself on a hard surface come the end. Anyway... if you have a complex window recess that you want to boolean into the wall then its probably easier if you do the boolean once into a local bit of wall and clean all that up leaving you with a nice rectangular piece of wall around the window then use that as your boolean object for more windows. The nice clean bit of wall is a lot easier to clean up than complex bits of window frames. Also, the boolean algorithms work better when there is nice clean orthogonal bits of polygon to do intersection calculations with compared to shards of rounded polygon bits etc.

[Robbie]

Redrobes, turns out Blender's boolean is actually quite clean. There's two solvers in the modifier, bmesh and carve...I got better results using carve, which only disrupted visible edges. I cut all the windows in one single operation and spent about ten minutes cleaning up errant vertices by snapping them to their most logical neighbor in the topology and running a remove duplicate vertices operation, which welds them cleanly.

Once I detach the exterior walls from the interior walls, I'll be doing some smart subdividing in order to get nice clean even spread of polygons to prevent long range culling issues, but I believe the engine handles culling pretty well...I'll definitely be discussing it with the Godot community for sure though. Thanks for the breakdown!

I didn't get as much time to work tonight as I would have liked, but here's an updated. I only really modeled one ward, but it was easy to duplicate it and get almost this whole end of the building.

3dWIP_04.jpg

And in order to see the character of the building start to take shape, here's a duplicated wing section up.

3dWIP_05.jpg

Once I add floors and ceilings, and of course the space between them, there'll be a bit more space between the levels vertically. Moving right along though!!!

[Robbie]

Just got a few minutes to throw in some floors and ceilings. Keep in mind, I'm doing this work on just one story and duplicating it before rendering these out. Before doing a proper topology consideration of the walls and all the cuts, one story of one ward is 704 faces. Not too bad. I'm shooting for about 1000 faces per ward for walls/ceilings/floors only. Poly count will of course go up once decorations are considered.

3dWIP_06.jpg

[Mouse]

You have cavities in the floors and the walls. Is that so the game engine can keep each room in its memory as a separate entity, or is it more to do with being authentic?

Its looking very impressive. I can almost see the textures and the landscape around it

[Robbie]

From what I've seen, with global illumination, the floors and ceilings need to not be continuous. It creates unrealistic light leaks as the resolution of the baked illumination may vary from system to system. If each surface is non continuous then the light will terminate on the edges as well. I'm not 100 percent sure though, but I'm hoping I'm planning it correctly. It is more authentic that way and I get texture space for each room if I want it. I'm really liking the character of the building though as it's coming through.

Sent from my HTC6535LVW using Tapatalk