This is a constantly difficult question, as over the years the workflow involved has changed a bit…
But the concept is:
- Enclose your scene in a cube or similar, simple poly shape.
- Fill this cube or shape with a parti_volume shader, which is “Participating Media Volume Shader”, in the long form.
- Attach lights and a physical light shader to the volume’s shape.
- Tweak for about two hours until it starts to look right.
I imagine after a few tries Step 4 would grow lower and lower, but it’s always a pain to do volumes in MR. The benefits though are that they look amazing, and combined with GI and caustics are very accurate.
The reason for Step 1 is that containing the calculations make them work faster; mental ray’s volume mathematics don’t have to work with infinite space, and therefore the math is much simpler and quicker to render. Most other software packages like you posted have built-in limiters that limit the math, but the negative to that approach is that you can’t surpass those limitations. MR starts with none and lets you build your own limits to suit a particular scene, which is (among other reasons) why it’s so complex to do volumetrics with mental ray.









