We often focus on whether we can do something, rather than whether we should. This Godot-made simulation is so impressive, you might forget for a moment it’s sick in every way.
As Bazyleeshek explained, there’s a camera using an orthographic projection and a low render distance, facing up to detect when something interacts with the goo. It takes the depth texture and renders it to a low-resolution viewport texture.
Two viewports, each watching the other with a one-frame delay, run the simulation based on the depth texture and the previous frame simulation result. The simulation itself involves four shaders in total. The final output is sampled by the goo material to offset vertices, warp UVs, and blend multiple textures based on depth.
The camera only covers a small area, but it snaps back to follow the player, adjusting the simulation coordinates accordingly if the player moves too far. This snapping is subtle, causing only a slight jitter, allowing the simulation area to feel virtually infinite with minimal performance impact.
Everything runs on the GPU via shaders, but the developer plans to optimize the goo mesh so that it maintains high detail only within the active simulation area.
Bazyleeshek also plans to do a full video breakdown at some point and even turn this into a real game. You can check out the original Reddit post for more info, and while you’re at it, take a look at their other game project about pets:
