Hi,
I hope this post isn’t too long.
I’ve uploaded a WIP to…
http://www.stalkingtheelusivepilky.net/ (5.5 MBs)
The terrain consists of 32258 polygons, a 2048 x 2048 texture, a diffuse detail shader, and a mesh collider.
The inverted terrain consists of 7938 polygons, a 2048 x 2048 texture, a diffuse detail shader, and a mesh collider (the latter necessary for some things I’ve tried).
The terrain is bound by a 12 polygon unrendered C4D cube imported with a collider (so the fps controller doesn’t fall off the terrain).
The larger C4D cube consists of 12 polygons, the same texture as the inverted terrain, a self-illuminated diffuse material, and no collider.
Thanks to a post I read I made sure all the geometry uses one-sided polys.
The GUIText is updated by a “for loop” that at most will loop 150 times before exiting, but immediately after loading loops just once, tests true, and exits (because the fps controller begins within the first bounds: see next sentence). The terrain is “covered” by 150 bounds stored in an array, and the aforementioned “for loop” tests to determine which bounds contains the fps controller and, if different from before, updates the GUIText with another snippet of text from an array of strings.
Another script keeps a point light above the fps controller at a height of 75 units.
So far, that’s all there is to it. I’ve tried many different things, including trees, water, line renderers, a sphere zipping around to allow the reading of the stories, particle systems, stained glass windows… What I’ve uploaded represents the only stuff I’m sure of (and I’m not sure about being able to wander up 90 degree slopes).
The terrain and its “ceiling” are created by lofting the photograph into polygons using Bryce then texturing the terrain with the photo.
I have a 1.67 GHz PPC Powerbook with 1.5 MBs RAM and a 128 MB ATI 9700 GPU. It seems to me that the thing is close to starting to chug along. If I add another 512 x 512 texture the framerate difference is noticeable.
So…any performance suggestions or other thoughts would be appreciated. For example, a post I read elsewhere in the forums made me wonder whether busting the terrain into smaller sections would improve performance (and maybe even allow me to up the poly count for greater, smoother detail).
Everything about this centers around the photograph and stories got by interviewing the two eldest females in it (I’m still working on the “stories”). So I’ve been judging anything I add by whether it serves the photograph. Although adding a tree model might seem to go with the photograph, it struck me as kind of a lame thing to do, not really adding anything. What seems more appropriate is to play with aspects of photography, like glass effects, lighting, questions about photography, etc.
I can’t claim to be 100% happy with the thing so far. I could use a poke in the head.
I’m very, very new to this 3D stuff and am excited about it.
Thanks,
Rick
Good start, and good luck!
pretty funky. it ran fine for me on a 733 mhz machine. i wouldn’t worry much on poly count. that’s rarely the bottle neck these days, especially at that count. 30K might be on the high side for that terrain but usually it’s materials and object count that cause too many render passes. textures, audio and movies take the most memory. you might look at the layered terrain shader approach instead of 2048 textures. tiling will give you much cleaner textures and you can use 512s (or even less) to save memory. interesting approach with the quadrant thing. not sure if triggers would have been easier but i guess what you did is more adaptable to what you’re doing. either way, it works! art-wise, if the game’s centered around the photo, maybe make it a bigger part of the game. there’s only certain clear views of it and it just kind of felt like some weird background. anyway, obviously it’s a wip - keep going!