Art pipeline and other things.

Art pipeline update

Greetings.

Ok, so more stuff on the Art pipeline. This week I spent time mulling over how various DCC apps export differently to FBX format. It seems there really isnt a single convention as to how one app saves as compared to another. On the upside though, the main advocate of FBX format also produces most of the main FBX content production tools (max, maya, motionbuilder) so I'm sure they will internally link up at some point and come to an understanding.

FBX format and the SDK is a tad weird though, because it has to support so many DCC tool features.

Originally I had envisioned using the FBX SDK from discreet for supporting Collada format files too (in addition to the obj/3ds formats it also handles). But having reviewed the code, I've decided to go for Feeling Software's FCollada libraries for it. FCollada is actually reasonably sane, if completely undocumented and frankly makes my eyes water less than the collada DOM and other api's I've seen floating around.

Right now I can read FBX format files with >1 UV sets (which is why I originally decided to do this), interestingly I was messing around with multi-sub materials in max trying to figure out how the FBX file would look internally after exporting a number of different setup scenarios (mostly in this case, adding a texture mapped box then trying to add different UV mappings and layers and textures and seeing how they worked). I've now figured out at least one useful workflow that seems to be what artists expect to do, so we're good to go from the FBX format front for at least moving forward.

So, next task is to get the DTSSDKPLUS to export the FBX format data from my internal scenegraph into Torque's DTS mesh format. From there its a matter of then adding the features to the Torque side to handle the various functionality I want (multiple UV sets etc) and then writing out the new features from the DTSSDKPLUS side and handling them in torque.

Once all of that works, I can integrate the Collada loader code into my scenegraph and the export should just work as-is. Same with .x and any other format I choose.

Game related

With a burst of productive energy right now, I'm having fun getting on with our games projects. Right now we're revamping the GUI to look more consistant and we're adding the UI to incorporate our shop-store backend system. I've come to the conclusion that writing custom GUI controls is for the birds and I'm using a compositional model a lot more (where you add controls to a GUIStackCtrl nested within a GUIScrollCtrl). Only downside is the pretty mixed up positioning elements and what appears to be a bug with extents of GUI controls not being updated properly when using the "width" and "height" of the parent control. Not to worry though.

I also started prototyping a simple 2D game during this weeks GID weekend. Although technically I wasnt GID'ing, it certainly kind of went that way. I implemented a simple heirarchy of objects in script, only to realize that I think I prefer to do that in C++ (plus the script was VEEEERY slow).

The basic premise of the game remains a closely guarded secret right now (ok, not closely, but I havent told anyone yet), but it involves a randomly generated heirarchy. The cool thing about using a randomly generated environment is that it allows me to offer unlimited mission possibilities.

Just to explain the random generation thing in a bit more detail. I decided to use a technique we used in the old worms games. When you generate a new game level, you are given an N character string. That string is the parameters to the generation algorithm encoded such that ANYONE can generate the same level based on that string. So sharing a good game level with someone else will be as simple as sharing 16 characters. Not only that but you can also randomly generate different TYPES of level based on different properties of the string (perhaps different bits in random parts of the string).

When players play the game, it is basically a logical puzzle, but it is bounded by time. You will earn a rating for any given level that gives you a "score" so to speak. So we can have people generate levels, have people DESIGN levels based on tweaking parameters, we can have people SHARE levels based on simple strings and we can SCORE players based on thier achievements for a given level. All with minimal effort and almost no data transfer. Sounds good no?

Another great advantage of this game idea, is that it essentially contains only procedurally generated content. Even the graphics will for the most part be procedurally generated (using an external application). In fact, the only thing that wont be generated is the world map. I'm hoping to find a source that will allow me to download the data I need to recreate that map though, so again NO artwork!

I'm happy with my prototype though. It feels like a good thing to do.