Apart from L-Systems, what are other methods of procedural tree generation that produce some nice results? Anyone?

There are also a number of other methods for procedural generation that dont use L-Systems..I have read a number of papers on such things. One method involves tracing the 3D paths of a series of particles, with gravity, wind, randomness, etc, and making the branches from these paths...If you provided for collision detection and avoidance, I'd imagine this could work very well.

Also, there are image-based methods that attempt to create an acceptable mesh from a photo or series of photos...such things exist in research papers and test implementations, I have not seen anything like this available publicly.

-----------------------------------------------------------

Depending on how "procedural" you want, you can make a number of prefab parts and write a script that assembles them together in a tree-like fashion.

I would use for example, a large trunk, big branches, small branches and leaves. Just make sure each model has a the origin point at the best connection point, and then somehow define some nodes for each model where the next piece will be attached.

Then you can assemble, scale, rotate, etc each part using parameters and so on. I have written such a program and it worked fairly well after I spent a while tweaking the rules.

One of my lecturers from my MSc days developed a 'strand' system for generating trees:

Strands, Gravity and Botanical Tree Imagery by Matthew Holton

http://www.gamedev.net/community/forums/topic.asp?topic_id=509303&whichpage=1�

Check my reply in that thread for a good method.