Impossible to resize voxel grid without breaking simulation


can something be done about the sim size settings? Currently, it’s borderline impossible to change then without breaking the simulation.

For example I would like to open a one of the example preset files bundled with embergen and use that as a starting point, but reduce excessive voxel density. No matter which parameters are touched, the voxel grid size always changes and completely breaks the preset example simulation.

I don’t know what kind of use case sim size settings currently accomodate for, but one of the most common use cases is the one where user wants to adjust voxel density without affecting the simulation grid size, and sim size settings in Embergen certainly fail to accomodate for that.

Right now there isn’t a voxel density setting to lower the resolution. That’s just not how EmberGen works right now. By breaking the simulation what do you mean? Hitting the bounds?


here’s a typical example. I open one of the presets, I really like it, and I would like to start tweaking it, but first I want to reduce the excessive simulation resolution for both performance reasons and also simply because I do not need such an extreme detail.

When I look at the grid settings, I see that the volume grid has specific dimensions. When I change the voxel size, the dimensions change, but I want to preserve the simulation, so I try to manually, in quite frustrating manner, set them back to the original values, but that is no longer possible because the dimensions are snapping to some very large increments.

And even when I am able to get very close to original grid dimensions after increasing the voxel size, more often than not, I end up breaking the simulation in the process, as shown in the video below:

It’s usually about 50:50 coin toss if I am able to change the simulation grid properties and maintain at least something similar to original result or not. It would be much more appropriate, logical, and user friendly, if there were only two sets of parameters:

  1. Grid size dimensions, but not in min-max pairs, just 3 float components of a single scale vector defining grid width, depth and height. There’s literally no benefit in having grid dimensions defined in min/max pairs, only poor usability.
  2. Voxel density. Same as voxel size, except when changed, it would not modify the grid dimensions, aside from rounding them up to closes voxel grid increment.
    Then the voxel counts, the amount of voxels for X Y and Z axes would be calculated from these parameters.

Right now, unfortunately, the grid setup is interdependency hell.