The idea is this. I have ground tiles and environmental props rendered in 2D quads, while characters in 3D. The choice to use 3D for characters (e.g. Total Influence Online, FOnline, Temple of Elemental Evil, etc) is to reduce the sprite work which grows exponentially when you want to show equipped items on the character. Additionally, making it a 3D model would give me much more flexibility in adding more animations, as well as some time savers like animation retargeting and blending.
Next, I'd love to try my hands on per-pixel lighting in this scene. This is the general high-level idea:
1) Render all diffuse maps into a render targets
2) Render all normal maps in world space into another render target
3) Depth sort all pixels by height and write into depth buffer
4) Use a shader that samples all render targets to figure out light intensity
Basically, a deferred rendering technique without polygons. Or so I read.
Thing is, I have completely no idea on how to achieve this. Seriously. I know how to do a full 3D scene, and I've written my own top-down tiled 2D game too. I understand 2D isometric draw orders and such, but right now I'll be using Irrlicht to attempt this engine.
This thread will hold my journey of discovery as I stumble and fumble through the dark. I appreciate all help I can get from experienced members here.
Here's the mockups that I did in photoshop to illustrate my eventual goal:
Day Time
![Image](http://splinteredcore.com/web/wp-content/uploads/2010/09/environment-mockup.jpg)
Night Time
![Image](http://splinteredcore.com/web/wp-content/uploads/2010/09/environment-mockup-2.jpg)