Happyday, on 11 August 2011 - 11:29 PM, said:
First of all, thanks a lot Frankie!
I've got two questions:
- Did I get this right, you're first lighting the map and then applying the textures?
- It seems like the map is build with brushes. That reminded me of a post by BladeKiller a while ago, suggesting to build whole map as ase models (link). Any comments on that? Is there one approach superior over the other?
Lets answer the two questions with one.
My current experience level with gtkradiant = 0
My current experience level with 3ds Max = My Kung Fu is strong.
That being said every building and model you saw in the video was done using 3ds Max and exported to ase and then imported into gtk.
Inside of 3ds Max you “must†first establish relative scale by laying out the map using proxies that represents the total volume of the area in that you wish to cover and what works well is to use real world values of 1 = 1 inch and then add 10%.
Once I had everything laid out in world space I then exported each of the individual components to be used as an x-reference and worked on each of the model as an individual file. SOP for building architecture based scenes, as you don’t run into data overload or editor lag problems. Once your ready to render you only need to x-reference in all of the individual elements along with their lighting rig and your good to go.
In this case.
I do a pre lighting rig on the proxy to get an over all sense of the of lighting I want and since normal and spec maps are light reactive it’s a good idea to have some kind of lighting in place to make sure that the output levels are not higher than they need to be.
I then hunt down the textures I need and apply them in creative ways then use Crazy Bump you make a normal/bump and spec map from the image. Crazy Bump for this purpose is ok to use as a place holder but unless you are reallllly happy with the result should be used as more of a place holder as more complex maps can add a lot of noise.
Of course at the same time I add the texture path name as required and copy the master textures to the game directory
What I have then is a single lego block that I already know how it will look as I now had a bit of experiences of what the map will look like in relationship to the completed model. This of course is where the experience comes in and no there is no superior approach over another just a predictable result.
In GTK I then import the lego blocks and set up the clipping that BK mentioned and the lighting elements as per the reference I did in 3ds Max. Since the clipping will always match the model, in most cases, I then only need to update the x-reference model and do an update witch updates all of the other elements that is x-referenced back to the other models.
As for numbers I’m still in the process in figuring out what’s going on but BK is for sure proven right that the performance levels between the two the all ase only map has a lot of the advantages and the current map in the video achieves a full 126 fps rock solid (screwed up by screen capture software) and the map compile time is twenty minutes.
Personally I don’t know if this is normal or not but a lot of my conversations with BK usually start with “that’s strange†;)
doing "stuff" with dead things.