How do you wrap a 2D image around a sphere? Well, the most familiar way to me is what 3ds max calls spherical mapping; you’ll have seen this if you’ve looked at a 2D atlas. The top and bottom poles of the sphere get a much larger area of the image than the middle of the sphere. Without correcting for this in some way, the top and bottom seem very pinched. I found a way of correcting for this, for a (horizontally tileable) texture in Photoshop. You can download a little action I made to help with this here. I’ll try and illustrate what you get from it with an image:
Top left – front view of sphere’s geometry with yellow UV seams at poles; top right – UV layout; bottom left – a cheap 2D grid texture I photoshopped together using my action; bottom right: the resultant sphere with texture applied. Although the grid on top and on the sides doesn’t match up, at least the scale is relatively consistent, which for most textures will be the most obvious thing if standing out.
Another way of mapping spheres, with less distortion, is using a cube map. This is just 6 images representing the positive and negative of the 3 axes – front, back, left, right, top, bottom. I’m probably going to use this method for a static “spherical” environment of a star-field, to indicate far off, where there is no geometry. It’s very cheap on modern graphics cards, and you can generate them in 3ds max easily using an existing environment. Here’s a quick test I did in the engine:
The cube map starfield was created in about a minute using a pflow system in max. It is apparently possible to generate a “real-time” cube map starfield texture procedurally, but I doubt my skills will be up to it in the near future. I’ll probably just create something prettier in max and photoshop, or maybe not use cube maps at all and go with pure geometry.
In the above image, the planet is a geosphere (triangles and no geometric poles) with spherical mapping using the method above with a PS generated texture. Obviously the lighting and shading are really dull at the moment. The vertical poles look pretty good, except when the LOD mesh kicks in – probably a UV interpolation issue. Not yet sure whether to go with procedural planets, moons etc, or bitmaps. I’ll leave it to when I really want to overhaul the graphics!
I’ve been using Ogre for a few days now, just going through the tutorials on the website and messing around, and I really like it. After the pain of programming a uni project directly in OpenGL, I was kind of dreading how much work it would be just to get a model into a real-time renderer. As an analogy, imagine wanting to cook a nice meal, and having to build the oven first – you know what you want the oven to be able to do, but actually going about it is a slow and painful process.
However, with Ogre 3D, the oven is built for you! If I don’t want to, I never need to go into quaternion maths, building a scene hierarchy system, frame listeners or any of the other complex engine necessities. After years of using 3dsmax, I found the Ogre workflow really simple and straightforward. There are lots of nice features and it seems robust. It is not too bloated in my opinion either – although the set-up process was rather long. The tutorials on the wiki website are really easy to read and learn from, although one was unable to compile on my computer. http://www.ogre3d.org/tikiwiki/tiki-index.php?page=Tutorials
Here are some screenshots of the demo scene I just made (having done Tutorials 1-6), using some example media and my own ships/ space-station:
You can read the code I wrote to create the scene shown in the screenshots here: http://www.evernote.com/shard/s121/sh/58d7c731-1507-4e4d-9e3d-8c9b14ef2005/71b5b5d250ca72c7173b8478e5b66395 – there’s exponential fog, texture shadows, a spotlight, a skydome, and some C++ structures like for loops to get me back into the flow! If you know some C++ and are somewhat familiar with making 3D graphics, I think the code’s pretty short and easy to read.
The structure of getting models into the engine takes some getting used to. There is a human-readable .scene file (which can be exported from Max, Maya, Blender, etc) and a dotSceneParser class. What this means is that scenes can be altered from within the .scene file, .mesh files and .material files without you having to recompile the source code. It’s amazing that you can use Max as your game editor almost, placing lights and objects, without having to constantly recompile your C++ code!
Here’s a screenshot of the “Easy Ogre Exporter” open-source Max plugin, http://www.ogre3d.org/tikiwiki/Easy+Ogre+Exporter, which is a really nice and simple way to get max scenes into the .scene format:
I had a really enjoyable few days taking it slow and learning all about Ogre – I don’t want to get disenfranchised by jumping in to making the game too quickly and doing everything wrong. Also I bought a load of expensive computer parts to make a new system, which will speed some things up a lot compared to working on my laptop – compiling and running the game takes about a minute, and those minutes sure add up. Hopefully I’ll start making the actual game soon though.