This is a scene that I’ve setup for testing purposes and it’s part of my research for various rendering engines, pipelines and work-flows. At first I wanted to model fully more than one room but I didn’t have that much time.
I’m trying to find and optimal work-flow (focused on Architectural Visualizations) that could be based to FOSS (as much as possible), from start to finish.
The part of modeling done by me was done in Blender.
Some textures were modified in GIMP.
Lighting done with one Sun Lamp, one HDRI Environment and Portals placed on every window.
For the materials I used a couple of custom nodes (based on older PBR tutorials by Cynicat Pro).
Rendered in Cycles with 4096 samples, about 1 hour and half per image 1280x720, on GTX 970.
DOF in Cycles.
Post Production is done in Blender.
This pipeline that uses Blender almost exclusively has many benefits.
You don’t have to move between different applications, Cycles is great for tweaking fast you materials and lighting and its compositing system worked without problems for me.
The PBR nodes gave me predictable results on every case and made modifications and testing quite fast and easy.
Contrary I still can’t get read from noise (though it doesn’t bother me that much in this specific case).
I would for sure prefer to have an artistic choice of the noise amount.
I’m hopping that Lucas StocknerDe-noising Builds will solve those problems in the near future.
(My testing with these builds so far are very promising, though buggy in complex scenes like this one).
Also I feel that 1 hour and a half is too much for this kind of images.
I’m usually asked to deliver images more than 3000px and that would mean approximately more than six hours per image on my current set-up.
On tight dead-lines this is kind of tough. Of course I could use a render farm or buy more GPU power (but that’s a different story).
Also on Post Production I’m planning to test Darktable and Natron which are both very capable and gave me great results so far.
Creative C&C are always more than welcomed.
Yafaray version here