Shadow of the colossus terrain...
Author
Check this article out, very cool
http://www.dyingduck.com/sotc/making_of_sotc.html
Im really curious about their whole terrain implementation. From what I can tell, the entire world is broken into chunks, and each chunk has three levels of detail. For the distant background, they render the 'super low' chunks as a texture.
My question is, what do they do with this texture? If theyre rendering to a texture, that means they have to be re using the same texture over multiple frames (otherwise it wouldnt be worth the effort). So does this mean they arent just rendering the current view, but a actual skybox (all sides surroundign the players current location). Then they only update this 'sky box' when the camera moves?
And how in gods name do they get the 'medium' detail models to blend into the background (a plain with a texture)? Would a simple radial vertex alpha fade work? It doesnt seem like the two layers would match up...
Any insight appreciated
I think the effect is called "impostors". If you go a little digging around here at gamedev I am sure you will bring up something on it...
I don't think this is regular imposters - from the looks of it the 'super low' detail level is actually a low-res skybox that gets updated only every few frames (and provides some cheap depth of field).
I'm not entirely sure how they stop the poping though (I've only played Collosus once) I suspect they aren't doing anything fancy but just expect the actual visible change to be minimal.
I'm not entirely sure how they stop the poping though (I've only played Collosus once) I suspect they aren't doing anything fancy but just expect the actual visible change to be minimal.
Well, if they took the medium detail mesh and used that to create the impostor, there would be minimal popping between the imposter and medium detail mesh stage.
I guess I haven't played the game, so I haven't seen the effect. I know they did something years ago with a game called Trespasser. They used impostors for very far away parts of the scene. They would update them every few frames so that the perspective was still almost correct. It sounds to me that they are doing something similar.
I guess I haven't played the game, so I haven't seen the effect. I know they did something years ago with a game called Trespasser. They used impostors for very far away parts of the scene. They would update them every few frames so that the perspective was still almost correct. It sounds to me that they are doing something similar.
I agree with orangy, I dont think they use imposters in the traditional sense, but rather take a snapshot of 8 fullscreen imposters, and slap them onto a skybox.
The thing I cant wrap my head around, is how do they do the PERFECT (it really looks perfect if you play the game) transition from skybox to 'low detail models'.
Like I said, will this work if you just render the skybox, then render the lowdetail models with a radial vertex alpha fade around the camera?
The thing I cant wrap my head around, is how do they do the PERFECT (it really looks perfect if you play the game) transition from skybox to 'low detail models'.
Like I said, will this work if you just render the skybox, then render the lowdetail models with a radial vertex alpha fade around the camera?
Seems to me to be "perfect" you just need to be consistent. Ie, the imposters are a real rendering with the same rendering system.. just taken with a 90 degree FOV camera pointed in axis aligned directions and placed into textures, which is mapped to a perecisely sized axis aligned box (not skybox!) that surrounds the axis aligned non-impostered geometry.
Probably each frame only one imposter texture is rendered (thus the speed savings.. 5:1)
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
