[Bf-committers] Google Summer of Code Idea

[Jean-Luc Peurière]

Le 25 avr. 06 à 02:38, Scott Johnson a écrit :

> That seems reasonable to me.  Are there specific shaders people  
> would like to see?  I was thinking perhaps a shader for  
> environmental effects (i.e. wind erosion, metal scrapes, dirt,  
> snow, etc...), along with a generator to generate random "dirty"  
> effects, as well as possibly a shader for reflections off of water/ 
> liquid surfaces.  I want to get familiar with the GLSL, so  
> implementing shaders such as these in Blender would be ideal for  
> what I want to do.  Are there others that are in dire need?

I have several remarks on this proposal :

- first i agree with Brecht that 3 months for just one shader seems  
quite a bit. now if several shaders are implemented that is ok.
- although very nice shader, usefulness is still narrow.
- if you are not familiar already with the blender code, doing it  
both for the renderer and the game-engine means learning
   2 very differents codebases. that seems to me not the best use of  
ressources, better focus on one and do more. In that case,
   that should be the renderer imho.  They are even not in same  
language.
- blender is only GL1.2, only the game engine support GLSL.
- A very important point is where does that features fit ? imho this  
is more than what the regular materials provide, but if you
   do it in the nodes, you have a problem with the GUI. Imho there is  
room for specialized shaders replacing completly the
   normal materials (eg great for NPR), but then you need to design a  
full easily extendable API. this is already much more
   work which must be accounted for.
- Along same lines, once that API exist, making it plug-in friendly  
would ease work for other renderers integration (probably too
   wide scope for SOC though).
- 5 weeks out of 12 on research & documentation is too much.

That said, i like this proposal, but you will need to refine it.

For other specialized shaders, you have 2 way :

* Photo-realistics ones, the tricky part is that most can either be  
simulated with regular materials, or wont fit easily in the render
engine (eg snow means displacement and mesh changes which means it is  
probably not a shader, but a modifier). Specialized
  shaders should be reserved to cases that cannot be done with  
regular materials imho.

* NPR are probably easier and a very useful feature, blender toon  
shader is very old.

the most advanced painterly NPR requires particules though.


lukep
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://projects.blender.org/pipermail/bf-committers/attachments/20060425/e0a60f26/attachment.htm