Some theory questions i have, looking at this problem.
You do have cylinder with applied sky texture imitating visible through window picture and light coming (FOR THE INSIDE SHOT) from out through window openings - Portals (whatever that might be), which actually is light source for your scene. Great. And i wont go over that while you move camera inside, window view will change just slightly…
At the same time software dealing with render task is plain stupid (i wont believe if you tell me different) and will go crazy trying to light, shadow, texture, represent, occlude and what not OUTSIDE of your model grabbing for that valuable CPU cycles.
Am i right assuming this?
Can we eliminate that?
Can we concentrate all effort and available resources only on inside calculations?
This model was for 100% percent imported in blender from another modeling software. See naming, texturing details in outliner. Fine.
For a still image meant to be rendered in Cycles which will build BVH tree once to get known whats behind what, where to shoot photons and such. That shall take time, CPU cycles, tenths of seconds. Not so significant if it is one still image but will make a huge impact if it is one frame of 1000 and geometry is complex, like in this.
Therefore, do we need to eliminate all not needed geometry if we plan on image sequence, how to find/consider what’s not needed?
For example, wall behind the camera - there will be just some twice or 3 times bounced light rays hitting camera from that; or better yet knowing that rays for now are shot from camera to scene…
Do we still need to keep that wall in scene and make calculations for this?
Any authorities out there, any comments, please!