| ||||||||||
| On Apr 19, 2018, at 10:41 PM, Chris Zwar wrote: > Sure it looks good but I don’t think it’s 100% mathematically perfect. I'm a little late to this thread, but... I don't think glow is inverse square, the apparent brightness of an object as it moves away from you in space is. If glow were simply inverse square, you could simulate the falloff by applying Levels with gamma 0.5 to your 0.0-1.0 linear falloff. Actual glow in the real world is a function of how many bounces the light does in the lens/air before getting to the film/CCD through a non-direct path. Depending on the circumstances, only a fraction of the light gets scattered in a certain direction, and then only a fraction of that is scattered back to the camera. Sounds more like a logarithmic falloff than inverse square. Not having a logarithmic falloff plug-in, the multiple blurs thing works pretty well for me. In line with the logarithmic thing, I usually do some sort of expression for glow radius like: glowRadiusStart * Math.pow(glowStep, glowLayerIndex) This gets big quickly, but that's the idea! And of course, to physically simulate glow you'd want a linear float color space with Add mode for each layer and the glowing object to be way over 1.0. Then because air/glass is pretty transparent I usually set the Transparency of each layer to 1% or so. Bonus points if you then bring the whole thing down a smidge (Levels output white 0.96 or something) because the glows added extra energy but the sum total of post-glow energy should be the same as the pre-glow. Brendan |
