Hi. My question is broad. I´ve been working on an idea that if I´m ever gonna actually build, it needs to be soon, fast and practical. I am planning on creating an animation where in a specific environment I´ve chosen (to provide human and relatable scale), a pile of human ragdolls start falling and piling up to show how the body count due to COVID progressed in my country (Brazil). As time advances, the death rate grew and the body pile too, and by the end we´d have a huge mountain of bodies which would finally make those numbers relatable to the average person. I pretend to show the animation alongside a running calendar/ date stamp and audio political statements made by our president downplaying the situation and undermining science and the pain of the victims and their families. It is important to me to get this done fast for we have elections this year and I´d like to contribute to stopping his reelection…
Thing is, my basic understanding of blender will get me into more trouble than results. So I´d like to ask for some pointers.
First: Is there anywhere I can get an average-sized human ragdoll model for free, that I can just add into the scene and worry only about dropping them on-queue? It really needs to be representative more than accurate for we´re talking about 677.000 dead overall. Simple geometry and maybe some articulation in arms and legs?
Second: What would be the most effective way, in your experience, of getting this rendered with minimal resources so that it is actually possible to render? I´m talking framerates, resolutions, etc. Considering that it´s intended for the internet and not TV. (cell phone screens, facebook, instagram, etc)
Third: Any tips and ideas at all that will point me in the right direction will be immensely welcome.
Thanks to everyone in advance!!!
One question:
What System do you work on?
Please post the specs. 677,000 rigged physics objects is a lot tho
I´m working on a PC running Windows 10
Intel(R) Core™ i5-9500 CPU @3.00GHz, 3000Mhz, 6 Cores
32GB RAM
It´s at work and doesn´t have a dedicated video card, but I could use a colleague´s PC for the rendering, which does have a video card for this purpuse. Nothing mind-blowing though
I know… I´m thinking that once the bodies fall on the ground they basically stop moving. And they don’t really show up all at once, although they do steadily increase… shouldn´t this make it lighter? Also, the bodies in the center of the pile could maybe be taken out of the model once they are no longer visible if this would help in any way
Your bottleneck is 100% going to be RAM. Physics simulations take a lot of RAM, and you’re going to breeze past 32 GB very quickly. Up until you run out of RAM, your CPU should be able to handle things smoothly, but once you run out of RAM, everything will freeze and probably shut down.
Not really- at some point in your animation, you’re going to have 677K active physics objects. The physics cache for one object can easily hit 500 MB, you would need 338500 GB of RAM in that case. Even if you can really simplify things and get the caches down to .5 MB, that’s still 338 GB of RAM.
Yeah, now you’re starting to get a good idea 
there’s no way that you can have all 677k objects visible and active at the same time, so what you should do is find out the maximum you can have, and then work on visual tricks so you don’t exceed that maximum.
Probably the easiest thing would be to either whip up a simple cube doll or use something like the default Mixamo mannequin. This video will show you the cube method:
Another thing you might do- use one ragdoll to represent 10 people. If you can squeeze the physic caches down to .5 MB, you’ll be sitting at 33.8 GB of RAM, which is still too much, but it’s a lot closer to what you can do. You’d still have 67K ragdolls. Honestly, I don’t think you’ll be able to see much of a difference between 67k and 677k objects visually, there’s just only so much you can have in frame and at that scale, an order of magnitude doesn’t mean much.
For example, here’s an array of 250k cubes with a 10 cm gap:
And here’s an array of 1 million cubes with a 10 cm gap:
See what I mean? There’s visually no difference, at all. Yes, this is a crude example, but orders of magnitude diminish visually proportionally to scale.
677.000 rigid coffins would be much lighter than the same amount of ragdoll people, right?
Yeah for sure, although rigid bodies are unpredictable and messy in high quantities. You can try it, something tells me it’s going to break badly once you get over 100k or so, but I could be wrong
I don’t think it’s super noob project, sorry. The technical requirements require some unusual and creative techniques. It is a cool idea though.
If I wanted to do this, I would probably end up combing rendered animations from many different files, so that I could use the same meshes as either rigid bodies or not rigid bodies. I would bake physics to keyframes so that I could use animated-type rigid bodies to span the breaks. In each new file, I would designate a non-rendering “pile” mesh as a passive colliding rigid body, which I would probably make by shrinkwrapping a plane to the existing pile and then using convex hull collision.
Anything involving 600k ragdolls-- at a minimum, 6 million different objects, and 4.2 million rigid body constraints-- is going to involve heavy scripting, if only to spawn the objects. So at that point, you pound out the process on the first few files so that you can script the entire process.