So I spent today converting the (30k triangle count) loading dock I showed in my previous entry into a lower polygon version...I was shooting to have it at 4k triangles,
I ended up using 3940...so I'm good haha. The 4k scene has more objects in it than the 30k one...unfortunatly I ended up having to essentially recreate the entire thing (except the concrete platform/stairs/railing).

Here's the end result...it looks almost the same as the previous scene just with a fraction of the polygon count.

Motion blur...


From up top...I'm loving the water shader :-)






So...yea I've got a little extra time for this entry so I'll post before & after screenshots from inside Milkshape3D in-case anybody is curious about how I dropped this scene from 30k triangles down to 4k with no real loss of data.

This is essentially how I create all my LOD models and also how I drop high poly count objects down to lower more usable counts (good to convert assets meant for pre-rendered scenes, into real-time models).



Here is what I started out with...my scene with 30k triangles...An entire city block in my game takes up about 3-4k triangles at it's highest level of detail...so 30k for a simple loading dock is out of the question. Also this entire scene uses a 1024x1024 texture (arranged as a 512x2048 to allow tiling of certain parts of the texture).

It was simple to convert the cargo containers, you just turn texturing on your Left-Top-Front viewports and grab a screenshot...then arrange the parts in your texture. It was a little more difficult for the crane itself because I had to isolate certain parts like the metal rigging, which needs to be transparent.

Here's a small version of the final texture, which again is 512x2048. The cargo containers use a seperate 512x512 texture each. You can see how only a 512x512 area is used for the crane, and the minimum necesassry parts were kept.


Once I have everything arranged in a texture I create a cube in the 3d model editor (mapped with the left-front-top views of the high resolution model) and essentially 'trace' the outline of each piece in each viewport...then combine all the piece and fill in the gaps when necessary.

I hope that was enough information, if anybody has anyquestions about this process, just let me know! You can really use this technique on any type of object.

Here's what I ended up with...the entire thing is < 4k triangles...which is good.


Traffic AI is still stupid...they know how to navigate the world, I just need to go through and get them to stay in their respective lanes. I had all of this working before...but I'm going to re-write all of the algorithms for stability purposes.


Random screenshot...


Ok that's all for now [grin].

- Dan