Custom Models[Adrenaline Syringe]

[xxxxxxxxxxxxxxxx]

Nice one Statyk!

You see, the thing is, when creating assets for a scene, you have to keep in mind how big every object will be on the screen in the end. Let's say your engine can render a maximum of 500.000 triangles on the screen when running on an average PC and you have enough memory for, say ten 2048x2048 textures. (or fourty 1024x1024 textures...you get the idea).

Now in reality there are of course dozens of other factors that determine how fast your scene will run in the end (number of lights, drawcalls per frame, texture filtering, antialiasing, postprocessing, how shadows are calculated, if you're using lightmaps or not, bla bla bla) But for simplicitys sake, lets only take a look at triangle count and texture sizes right now because those are the most important when you're modelling.

Now your goal is to make the scene look as good as possible, so obviously you need to spend your limited polygon and texture size budget there, where it is the most visible. That means that bigger and more important objects get more polygons and more texture space than smaller, less important ones. The same goes for objects that will be seen in close-ups versus those that are meant for background use.

On a character for example it is not too unusual to spend a third (or even more!) of the polygon budget on the face alone, because it is often seen close up and it is also the main point of interest, where the eyes of the viewer will be drawn automatically.
Or a weapon that is meant for first-person use may have twice the number of triangles than a weapon that is wielded by an enemy or only used in third person, because it is further away from the camera then.

So, to determine how many polygons you can spend on your syringe, you have to first ask yourself how it will be used. If its just an ambient object lying around somewhere, going for the minimum of triangles possible is the way to go. If its used in close-ups (Like, say, you have an animation where the character jabs it into his own arm or something along those lines) you may want to make it more detailled so it wont look out of place compared to other close up objects. (Like the lantern or the characters hands)

Last but not least, there are LODs, Levels-of-Detail. I don't know if and, if so, how HPL implements these, but in most game engines you have several versions of your model at the same time with increasing polycounts. (up to five) The most detailled version is used whenever the object is close to the camera, the lower versions are switched in automatically when the camera is moved further away.

Same is done with textures by the way, but I believe the .dds format creates mipmaps ("texture LODs") automatically. Only that with textures it is not only for performance reasons (although it helps performance as well of course) but also to avoid the Moiré - effect that would be visible when viewing a high-res texture from afar.... but my wall of text is long enough already, so I may talk about that topic some other time