Horde3D

http://www.gamedev.net/community/forums/topic.asp?topic_id=482962

This is an interesting idea - drawing water as a (deferred shading) post-processing effect instead of as geometry. He doesn't really explain his technique, but I'm guessing he's just sampling the position from the G-buffer and calculating the amount of water between that position and the camera.

Very cool!

I wonder if this technique could be improved by rendering a real plane for the water in world/view-space rather than a full screen-space quad. This way it would be cheaper to get the texture coordinates for the wave map (no reconstruction from the position buffer needed) and you also get clipping for free (only the visible water surface is drawn).

Hopefully he will give more details on his implementation...

Reading between the lines, I think he was trying to avoid this at all costs, to avoid the messy problems at the coast line.

I wouldn't be surprised if he gave the plane a shot though first as opposed to saying "Hey! What can I do in screen space? WATER!" However, being based on volumetric fog I can't see why you'd want to avoid a plane.

At the least I think a plane at the maximum wave height would be worthwhile even if it were just a super light set the "alpha so alpha test can filter out parts that don't need to be processed" pre-pass. But then masks are my personal universal solution to all problems, even the ones that don't exist.

EDIT: But the biggest question I have is what does it look like underwater?

Yeah, it's interesting that he didn't post any pictures of that

It should work fine in theory though, as he should be able to calculate the amount of water between the camera and a pixel, so when underwater, every pixel should return a positive value in that test.

The trouble is it LOOKS like fog. The image is striking at first glance but the surface patterning is not that convincing and at the edges it looks quite disconcerting because the water has no substance.

I think its more proof of concept and for his purposes than meant to overtake the world as the new standard way of doing water. Side note: I really like that "lack of substance," if the water weren't so opaque it would be considerably more convincing.

I think you're right, also looking at the shots on the website the surface rendering is quite nice really.

For anyone who hasn't seen it before there is some very nice real-time water rendering in the Typhoon engine. The site includes rough descriptions of the techniques used and refers to papers (not linked, just gives titles and I haven't tried to find them yet).

Also, looking at Typhoon again, the coast is the one area that doesn't look all that good. There the water ends in abruptly and the foggy approach suddenly doesn't look that bad by comparison.

Oh and another thing about that engine is it demonstrated to me the difference between of using streak and bloom filters together which feels like actually being in the place in the presence of intense light and just bloom which has a cinematic effect. i.e. the optics of the eye create streaks and film cameras don't so much.