Help with "2012" style fireballs

I need to make fireballs rain from the sky in the style of the lava bombs from the Yellowstone eruption in 2012. These will be composited into multiple aerial plates. I’ve got a pretty good handle on the fire part, but the smoke is looking too thin and too “soft”. I need the detail and texture of these:




Any ideas on settings I should look at? I’m getting close by increasing the divisions, but the baking time is growing with that too of course. I need to be able to animate the emitters, bake it, and send it to the farm as quick as possible since I have quite a few shots to do.


Meanwhile, here’s what I was getting with divisions at 128. If you click it and enlarge, you will see how “soft” they are, especially the ones on the right.

Fireball01-3-15.blend (724 KB)

Incase anyone wants to take a look, I’d be thankful.

Well, there is a reason why highly detailed simulations and destruction effects cost a lot of time and money. I’m not much of a simulation maker so can’t help with practical side. Any chance you could also bake your sims in farm? This way you could work on next shot while others are being baked.

Yes high division sims take ages to bake. Also you can probably get away with just simulating one then offset the frame start and rotate, move it into other locations.

That’s a good idea, I’ll have to try that

Update: Things are looking more promising. I had the crazy idea that if I kept my settings the same and scaled the emitter up 3x or 4x, then I should in theory have 3x or 4x as much detail in the smoke. Crazy idea seems to be working. I’ll still need to tweak things, but I’m on the right track!

Do you mean that the emitter object is larger? Will that ruin the effect at the hot point?

Here’s the test render from last night with my larger emitter

You can see it actually worked pretty well. 200 Divisions, and Hi-Res set to 3 with a strength of 5. The smoke has detail which is good, but now I need to figure out how to stop it from spreading as much and how to remove the noise. Doing another test now with with a strength of 2

Not if I move the camera further back to compensate


Setting the strength to 2 didn’t work. I ended up using the settings from last night’s render, and just turning down the vorticity from 2 to 1. That did the trick and now we’re good to go.

How long did your sim bake take?

For 200 frames, it took around 20ish minutes. For fireballs in the far background, I’ll need a larger domain so the smoke trails can be longer and will do those at maybe 100 Divisions instead of 200. That should help speed things up. I’m currently making a few generic “stock elements” that I can use, and then do custom, high res ones for close ups.

Been working on this for a LOOOOONNNGGG time and can’t seem to get it just the way the director wants it. We’re still trying to create a 4000x4000 stock element that has the fireball going across the screen leaving a fairly tight smoke trail, then the smoke trail slowly rises slightly and starts to dissipate over the course of 1000 frames.

If I bake the smoke around 150, the dissipation looks great but the fire and smoke look really soft and bubbly. If I bake it around 350, the fireball and the smoke look amazing, but the the dissipation gets all streaky.

I really need this to be able to work in one continuous shot given how it’ll be used. Any ideas, or does anyone want to take a crack at it? At this point, we’d be willing to pay if someone can get the look, animation and dissipation that we need, all in one shot.
Fireball01-5-7.blend (640.4 KB)

Does anyone know what “units” are used for the temperatures? I set Blender to Imperial but that doesn’t seem to make any changes as far as making the temperature values easier to understand.

Sorry but I don’t think that any of the physics stuff in Blender has real world or sensible units. All the tutorials I have seen tend to conclude with “I don’t know why but just set the slider to this value”. So its a LOT of trial and error. And when you’re baking all the time that is super frustrating.