Hello I''m actually working on a terrain generator/renderer and I encounter some troubles... In a tutorial i saw how to texture map an entire height map by stretching a texture over the complete terrain. Is there a better method to texture map a height map ? I''m getting some undesirable texture "slicing" when i move through the terrain like if the texture isn''t fitting correctly and of course i would appreciate to avoid this effect by using another method. Another problem i encounter is that i want to be able to render large height maps. I saw some algorithms like ROAM, CLOD, .. that allows to reduce significantly the number of primitives to draw each frame but can someone advise me in which one to use ? Finally my last question When I see nice demos like the one available on NeHe''s site i''m always pleased by some camera movements, i mean the way the camera is following a defined path. But i''m still wondering myself how it is possible to save predefined camera path efficiently. I mean, which way can we save a camera path without keeping tons of bytes about it and what kind of implementation will you recommend to get the best results ? In advance thank you for your help Best Regards, RaPhiuS ps : hope my english is not so bad =)

I am also searching for a good way to tassellate terrain. I came to a nice idea but i''m not too convinced about it. Furthermore, the amount of computation required looks way too high so I cannot really tell how it works before I try it out. I would be pleased to check out what other people is thinking.

I heard ROAM is the best but I took a look at it and it does not seems so good to me. It requires to tassellate geometry in a way I find difficult to predict. BTW, the algorightm itself is so generic that it could be used for everything - i feel referring to it as a terraing rendering algorithm is not really right.

All the lod-based things are somewhat problematic to me, but they look easier to manage, especially if you can access accelerated vertex arrays like ARB_VAO. The real problem I feel with them is the amount of vertex data they keep flying around.
Does CLOD mean Continous LOD? This is probably what I am going to implement in my terrain subsystem I''m not sure because I did all the algorithm design from scratch.

If I have to implement a camera, I feel the best thing is to have it follow/point to a spline. Lerping may also work - it''ll be easier to implement runs somewhat faster at the cost of much simplier movement.
My idea would be to have a path AB. We know the camera takes T seconds to go from A to B. To move along a linear path I just need to interpolate accordingly. Splining along ABC is more difficult but more effective.
Splining would greatly reduce the amount of bytes needed when compared to lerping, which is probably what you already figured out to do.

Bye