This scene was inspired by a “20 megabyte” challenge in which the entire packed .blend file for a scene is supposed to stay under 20 megabytes (maybe that was self explanatory).
While the version of the scene I used for these renders is closer to a gigabyte, the final scene for the challenge was about 16 megabytes (I swapped out the JPEG compressed textures for the original full-resolution TIFF).
Recently I have been on a bit of a nostalgia trip with Blender 2.64 and Luxrender 1.1 (both circa 2012). This scene was created and modeled in Blender 2.64, aside from the cherub, being a photoscan I created a couple years back using Meshroom and about 400 images taken with my phone. I was also perplexed by the fact that I had never achieved good results with Luxrender’s SPPM engine, and was motivated to finally make use of it after all these years.
I was inspired to add bubbles to the scene later, as I was curious of applications of Luxrender’s support for thin film interference. In Luxrender, I used the glass node with 256mm thin film, multiplied by distorted noise to add variance to the thickness. In LuxCoreRender, I created Gaussian pattern color noise in Photoshop, then used various blurs and hue adjustments to create a texture, which was mapped to the glass reflection color. The texture I created to add color can be seen below. This may be useful in Cycles and Eevee to simulate the effect!
Here is a comparison of the two scenes mostly converged. The Luxrender scene (Luxrender 1.1, Blender 2.64) ran for about 12 hours on an old machine with 2x Xeon X5690s (12 cores, 24 threads, ~3.5 GHz), the LuxCore scene (LuxCoreRender 2.3, Blender 2.82) ran for 1 hour on the same machine then was cleaned up by OIDN. Luxrender SPPM is the first image, LuxCore BiDir is the bottom image.
Interesting for sure, and the differences are mainly in the brightness and quantity of reflections and refractions. Overall, there is a certain charm to letting something render for 12 hours, only to find it is still noisy the next morning!
I am interested though, does anyone else purposely use any old pieces of software in their workflow? If so, why do you prefer the older version?