View Full Version : Poly Reduction with Terragen obj's
07 July 2003, 01:05 AM
OK thanks to Duke I have been playing with WorldMachine/Terragen and exporting the incredible landscapes as LW *.obj files. They transfer so nice and neat to LW it makes you cry. Why would I cry you may ask? Because of the couple hundred thousand polys it creates. So here my question what is the best way to reduce the poly count with the catch being not having to go in and start tweaking everwhere to clean the geometry back up. The two tools I was playing with were naturally qemLOSS2 and Reduce Polygons, but after a couple reductions the geometry goes non-manifold all over the place. Any tips/tricks, plugins that might help here?
07 July 2003, 01:59 AM
Hmmm... Well, I'm assuming your object gets imported as triangles. If that's the case, you might try Merge TrigonsX. It will convert triangles into quads (and thereby halfing the number of polys):
You could also try the latest version of qemLOSS, which appears to have many more features than version 2.
And for whatever it's worth, the normal mapping plugin might help to retain the look of your object, even on a severely low-poly model:
Hope this helps!
07 July 2003, 02:44 AM
Thanks for the script tips, Steve... Didn't know these, but they are very, very useful... Great timesavers ;-)
07 July 2003, 04:12 AM
Thanks for the plugin tips Steve. I just went and got both and will play with them in the morning. I'll let you know how it turns out.
07 July 2003, 10:32 PM
OK the best results I found was to use the NormalMapper plugins to bake the texture, which took more then a couple minutes to get the first time around. Then I ran the Merge_TrigonsX to cut the poly count in half. this provided the best results. If I then tried to run qemLOSS2 the whole object blew apart. If I ran qemLOSS and tweaked the values abit I got a decent reduction but lost alot of quality and still had to fix non-manifold polys left and right. So unless anybody else is used to playing with these high poly surfaces and has a few tips to share we'll just go for the halving and celebrate I can get that much and still keep almost all the quality.
07 July 2003, 10:57 PM
Aurora: Polygon reduction, especially when used in conjunction with Normal Mapping plugins like Microwave, Normal Mapper, etc. is really an interesting topic. Would you mind sharing renders of your results?
07 July 2003, 11:19 PM
You know what I'm tired of not having my website up. I think I'll grab me a free host and set something up so I can start posting images again. But its going to be a week or two before I get anything posted. I'm leaving on vacation in the morning and when I get back next week I have my err @)th high school reunion which I'm missing Siggraph for. How lame is that:annoyed: I'll let you know when I have something up to show.
As for Microwave man I love what you can do with that. I almost sacrificed getting a DVD writer to get Microwave instead. But NormalMapper is no slouch with what I have seen.
I'll tell you what I'll write a full tut on using WorldMachine to create a landscape, transferring it to terragen tweaken it exporting to LW. Once in LW texturing, using NormalMapper and MergeTrigonsX to create a somewhat kickass landscape. Hows that sound.
07 July 2003, 11:35 PM
:drool: <--- That's how it sounds. :)
07 July 2003, 09:57 PM
I'm not sure why you are getting reduction problems on your terrain when using qemLOSS. I reduce terrains all the time with it, and most of the time can easily reduce the terrain to at least 10% (many times even less) of it's original polygons, without losing important detail.
I only have a couple of guesses, and that is if the vertices of the Terragen object are not all merged, or if it does not have all the polygon normals aligned in a uniform direction. I would think the last guess is the most probable cause and the fact that they are not aligned would not be obvious if the double-sided surface attribute is turned on. Try doing an "Align" on the Polygons and see if any normals are flipped during that process.
If that fails, I would be glad to look at a sample object and see if I can figure out what the problem might be.
01 January 2006, 05:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.