- This topic has 18 replies, 4 voices, and was last updated 2 years, 9 months ago by Eric.
December 15, 2019 at 1:23 pm #4014Jeremy Hill
Awhile ago, Peter Shirley issued a little challenge:
Real photo. Ray tracing challenge pic.twitter.com/crmhaVoswD
— Peter Shirley 🔮🧢 (@Peter_shirley) November 29, 2019
So eventually I had to give it a try with bella:
@Peter_shirley here's an attempt.😀
Lighting is solely by the 9 mesh lights, and touch of IBL to give the color cast to the walls. Glass is a D-K9 crown.
Also a second, more appealing view of the scene (sans IBL). No post-work on either image. pic.twitter.com/laK8Ab8dhY
— bella (@BellaRender) December 14, 2019December 15, 2019 at 3:10 pm #4016Philip
PhilipDecember 17, 2019 at 7:08 pm #4026Thomas An.
Epic !December 20, 2019 at 10:15 pm #4054Eric
Sorry JD, I don’t think that looks much like the photo, nor much different from any other render engines output. The effect of glass glowing, due to surface imperfections or volume impurities, is not present.December 21, 2019 at 4:45 am #4055Jeremy Hill
I agree it is not a good match, but the test was more about this being quite a difficult lighting scenario, given that about all we see is from light refracted back out of the crystals.December 21, 2019 at 9:10 am #4057Thomas An.
I don’t agree with Eric. This was a tongue-in-cheek poorly defined challenge. It’s an iPhone photo from a random location and Shirley never told us what else was in the room contributing to the reflections. A proper challenge would have all variables controlled in a well defined environment. So, given how informal this was, Bella is doing an impressive job with dispersion and refraction.December 21, 2019 at 8:48 pm #4059Eric
I want an ‘sss’ type thingy which mimics glass glow, normal sss can’t do it. Glass glow matters!December 28, 2019 at 1:54 pm #4074Jeremy HillJanuary 1, 2020 at 7:22 am #4081Thomas An.
Impressive numbers but for the wrong reasons 🙂 This is one case where extra time didn’t matter. All essential light elements/phenomena were present in the first image. In which case a simple denoise would have sufficed.
I would even consider the bold move of hard limiting all renders to less than 10hr. The Bella engine should not allow you to enter higher times. If the engine can’t do something in a max of 10hr just do something else.January 1, 2020 at 2:42 pm #4082Jeremy Hill
Hmm, not sure about that idea, what about when I want to render something 10k x 10k pixels?January 1, 2020 at 7:32 pm #4083Thomas An.
Hmm, not sure about that idea, what about when I want to render something 10k x 10k pixels?
LoL, Confucius says “you should rid yourself of such wants to be free of pain”. If someone wanted big prints like that they’d consider a larger render farm to get to higher render level within the time frame.
The premise is that very few people on the planet (maybe five ?) today are actually willing to tie up resources for over 10hr on a single image. Back in the day it was curiosity. Today, not so much unless the engine is also solving the time-space-continuum.January 1, 2020 at 7:53 pm #4084Jeremy Hill
Probably both true, but I still don’t see a reason for artificially limiting it. An argument can be made for making the default render time limit to be 10 hours or such, but beyond that you will add complexity (logic in resume/merge/etc) for no (to my thinking) good reason.January 4, 2020 at 12:08 am #4090Eric
I think it would render a lot quicker with phong shading / round spheres, even small fillets might speed it up, there would be more geo but also more ray ‘cohesion’ around corners.. I’m just guessing.January 4, 2020 at 3:31 am #4091Jeremy Hill
It did not take that long to be acceptable, I was mainly just running it as an extreme test, to ensure that there is nowhere in the code that starts misbehaving once you get into dealing with very large numbers.
As for whether it would render faster with another shading model or other geometry, that may be possible, but I suppose a similar effect could be produced by just restricting the number of bounces, too.January 4, 2020 at 11:03 pm #4094Eric
I think the way to get glass to glow, is for a tiny piece of light to be emitted at the point of every internal bounce, such a tiny amount but it builds up, that might contradict contemporary science but they are always revising shit, so my hunch is that it’s an entirely plausible cause of ‘glass glow’.
Einstein was a fraud, just look at his grin.
You can put sss into just the LAYER, and turn on phong shading.. maybe works. SSS with pos assymetry I guess.
- You must be logged in to reply to this topic.