Le 2023-12-05 à 13:26, Chris R a écrit :
> Alain Martel <kua### [at] videotronca> wrote:
>> Le 2023-12-05 à 08:42, Chris R a écrit :
>>> I have gone back to working on poor Cogito in his cell.  In this new version, I
>>> think he has a lot more character due to adding more grime and rust to his body,
>>> and placing the rust in more strategic places.  The dark brown streakiness on
>>> his back and head are from accumulated grime mixed with dripping water from the
>>> ceiling.  I have placed a ring of rust around his eyes and mouth as those would
>>> be places where the metal had been cut and would be more susceptible to rusting.
>>>
>>> In general, all of his body has been un-shined.
>>>
>>> In addition, I did a fair amount of work on the walls of the cell.  The stones
>>> are now much rougher, and I added grime of many colors in a gradient from the
>>> floor to the ceiling.  There is some reflection as well where the walls are wet.
>>>
>>> There are some streams of dripping water from the ceiling as well.
>>>
>>> The floor was modified slightly, but I haven't done enough work making it as
>>> dirty as the walls, so that's probably the next task.
>>>
>>> This version was done without radiosity and with moderate anti-aliasing.  I
>>> tried rendering with radiosity and a higher degree of anti-aliasing and found
>>> that it took 8 hours to do the first radiosity pass, so I abandoned that.  I
>>> later realized that the two light bulbs in the scene (one in the cell, one
>>> behind the camera) were set up as 3x3 area lights, so the scene had 18+ light
>>> sources.  I put those back to single-lights and the render took about 24 hours,
>>> so I can probably add in radiosity again later.
>>>
>>> -- Chris R.
>>> -- Happily rendering since 2014
>>
>> 3x3 area_light is small. Do you use adaptive ?
>>
>> Use the same lights, increase the area_light to something like 17x17,
>> and use adaptive 0.
>>
>> Thar way, only 4 samples per light will be used for much of the scene.
>> Given the composition, the risk of having artifacts is pretty low.
> 
> Ah, thanks.  I had pulled the light source for the bulbs from the LightSys code
> which uses adaptive 1.  Using adaptive 0 it does seem to be rendering more
> quickly.  I took your suggestion of using 17x17, made it circular and used
> orient as well.
> 
> -- Chris R.
> -- Happily rendering since 2014
> 
> 
Here are the starting arrays for various adaptive values :
0	2x2	2^0+1
1	3x3	2^1+1
2	5x5	2^2+1
3	9x9	3^2+1
4	17x17	4^2+1

Or, 2^(adaptive value) +1

To get the benefit of the adaptive feature, you should have an array 
density at least two higher than the starting size of the adaptive value.
A 10x10 array will silently get bumped up to 17x17.

In my experience, an area_light 257x257, 513x513 or even 1025x1025 with 
adaptive 0, 1 or 2 is barely slower than a 9x9 one with the same 
adaptive values. And, those don't need any jitter to give very smooth 
results.

Post a reply to this message