by Zach Harvest
This week I worked on finalizing an image. Again, not a lot of content this week due to the nature of my progress so it should be short. I spent all week rendering and re-rendering and rendering and re...you get the picture. I ended up getting a actual shape that looked cool enough to focus on making it look good. I love the reflections of the line lights and went with an amber glass look. To contrast for color I use slightly blue hue lights so in edit I can pull the color I want. LOOKS SO COOL. This is nice because it doesn’t seem like you're wasting time rendering. Even when learning, if you have a shitty scene and image, it always sucks to spend for fucking ever rendering something that you don’t even like in the first place. I also got to drill into some other features I wanted to use like Volumetrics, de-noise image, and sample amounts. Their volumetrics work like most others but with changes in how far the constraints go. I had to play with the numbers for a good long while to get it looking the way I wanted. Side note. Renderman Volumes are the tits :D When working with volumetrics in the past, I never had any real tricky caustics to deal with in the volume. This reflective colored surface is FUCKING up my day with specular scatter n such... This meant I had to do a realllly long render. I’ve said this before, but I will say it again. FUCK these renders take forever. I mean, I keep adding filters...so I don’t know what I expected anyway. All of this stuff would have slowed down Cycles too, but damn... I will upload all of my progress shots which are in #renders anyway but I figured you guys should see the amount of renders I have gotten off trying to get everything looking the way I want. What a time consuming process...my “final” (probably not final) 1080p render took 25 fucking hours and it’s still a hot mess with noise. The nice thing about Renderman is you can just throw more samples at it, and that should fix it. I need a second computer for that however. I did a shorter 3 hour render with MUCH lower samples, we talking 300 vs 3000 and used denoise as part of Renderman and it does a pretty good fucking job. I might need to add some more samples like around 1000 but that saves me ⅔ render time if it looks good enough. But I have some things to work out, like how long it will take to actually finish. I’ve just been throwing samples at it until I see that it looks good enough then I cancel the render and save it. This has been really nice for testing but getting a “finished” render with post options means I need to let it complete. Also a value I didn’t know how it worked until this week is the minimum sample rate which looks and sees parts that no longer require samples so it can focus on the parts that do. The issue with this is sometimes the background remains noisy because “it doesn’t need more samples” usually this is nice and saves time but I got better renders out when I turned up that minimum value. So check out the lots of wips that all look similar lol.
Last Edited on Tue May 31 2016 14:35:21 GMT-0400 (EDT)
Sick! Can't speak for renderman specifically, but with some of these renderers they tend to reach a similar level of quality in similar times almost regardless of resolution (that's an exaggeration but roughly it can hold). One thing to try is comparing a 7hour render at 1080p with a 7 hour render at double that resolution and then scale down to 1080p. I had a friend doing this the other day and the down-sampling filtered out the grain surprisingly well even compared to more complex approaches.
on Tue May 31 2016 16:09:52 GMT-0400 (EDT)