FROM THE COMMUNITY COLLECTION
Shadow Experiment
Here are the results of my failed experiments to render Chantier with soft shadows. a: Sun only, hard shadows (1200 x 1600 pixels render time 1 3/4 hours). b: Sun only, soft shadows, rendered with 36 rays/pix; time to render 18 1/2 hours. c: A ring with 12 radials without falloff and inverse gradient at the place of the sun. d: Black HDRI with a bright sun, replacing Bryce sun. Rendered with 256 lights, but only 4 show up. e: HDRI subjected to an external median-cut algorithm creating 256 lights since the Bryce one gave only 4 lights. Rendered with 256 lights. f: Like e, but rendered with 1024 lights (quality=1024), i.e. 4 lights per median-cut dot.
Loading unique views3,600 unverified legacy hits0.0 archived rating2008-09-06 added
Comments 4
A conversation around the artworkSign in to join the conversation.
Depending on what type of light setup one is looking forward to achieve, every pic can make sense. As a "fan" of soft shadows I personally preffer "b" and "c". In all images there's this nice color gradient on the ground showing slightly reddish on the left to a bit bluish on the right. I guess that is all. I wonder where this is coming from...
C looks indeed the most natural but there is shadow banding. If I had used more radials, the banding might have disappeared. That's why I resorted to an HDRI, which was easier to create than all those lights. Something is still wrong with how Bryce applies the median-cut alg. Using an external one at least gives the results expected. B is not natural at all because the structure that casts the shadows is actually so small that it fits into the camera symbol and the sun is at "infinite" distance. As for the render time - it is from the large and complete renders. The other 4 renders were made at the size shown here, so I cannot give render times that are comparable.
Interesting. The default Bryce sun only has one soft shadow setting. This is unfortunate. This is just one reason I wish they would allow for a Radial Sun, that would create a sunlight that was controlled exactly like a radial, with the lack of falloff already assumed. This would allow you to use a slider to estimate the softening. So the question is, how soft should soft shadows actually be? The IBL sun is a very interesting thing to study because even though you had a black hdri with only a white spot for sunlight, it still failed to assign all 256 lights to the white spot. If this had worked as I would have expected you would have gotten a nice soft shadow sunlight from the hdri. I myself discovered this soft shadow problem when I was working on the Perfectville Scenes, finding it hard to create good soft shadows with the hdri hotspot. So I have abandoned the hot spot hdri idea for now and I am using a dome of 80-146 lights as soft shadow sun. At least this way I can always know exactly how many lights are being used as sunlight. I agree that "C" is the best looking if indeed the shadows are meant to have any form at all. Perhaps "B" is the most real with the extreme softening, it all depends on how wide the light source is supposed to be. By the way, "B" took a hell of a long time, over 18 hours! I think in this case you should have gone for standard AA and avoid the premium settings unless absolutely necessary.
. C looks best of course, but what was the render time on that?