Simplifying Cast Shadows

I have a semi complex model of a tree in a scene with several linked duplicates to create a forest setting. Fairly low res, but good enough for my purposes at the moment. My CPU is not the most powerful thing, and when rendering the scene, Blender crashes.

I’m wondering if there is any way to use a simplified shape to cast a shadow on the ground in the scene, that doesn’t actually render? So detailed tree models would show but the cast shadow would come from a transparent (or non displayed) blob. Thanks for any ideas!

NOTE: All of the following assumes you are using Cycles. If you aren’t, please say so.

Memory wise, (which is probably your issue) this doesn’t save you anything. You still have to load the direct-viewed tree AND the shadow tree. Remember that linked copies are only instances if they don’t have modifiers, you should use duplis instead for something you need instances for (like trees).

Speed-wise, you might be able to save some CPU time, but it would be hard. It’ll probably be tricky to get a shadow caster that looks right without resorting to alpha maps, and you don’t want to use those if you are trying to optimize shadows. Transparent shadows are significantly more expensive than solid shadows, even if the casting mesh is much lighter.

If you post more info on your scene and your computer specs, and maybe try hiding some objects to isolate what’s causing the crash, we can help you out a lot more.

I’m using the default rendering engine for blender 2.7. I think that’s cycles.

The slightly detailed trees become too demanding pretty fast for my 2.4 ghz duo core with only 3.5 gigs of ram. I’m primarily doing projects that are for some personal study, so the underpowered CPU isn’t the biggest concern.

Seems like casting shadows from branch to branch is requiring a lot of memory, but having a simple cone shaped tree works. Was looking to up the detail a little on trees but leave the shadow as only cast onto the ground. Wondered if that was something done in render layers or possibly something that could be done with transparent empties (can’t see how to get a cast shadow with those), or maybe something else?


You can definitely break the scene into smaller, simpler pieces that you then composite together … and, I would frankly suggest doing that, no matter how powerful your machine is (or isn’t).

Think about how folks put together audio-recordings: their “mixing boards” might be 80 or more channels wide, and they might use all of them. The (union scale …) musicians come in, do their parts, pick up their checks, and then leave. Now, the work of making the actual recording “from all that data” begins.

So, yes: there can be a “track” of information which merely tells you “where the shadow falls, and how intense the shadow is.” But that channel contains: “only the shadow.” Not its color, nothing but the shadow. Only where it is and how dense it is.

“In post,” then, you can do anything you need to do with that shadow. Computationally speaking, the process is now free-of-charge and you see your results instantly.