Follow the installation steps there. If you experience any issues, you can ask here or on GitHub.
The repository also contains example files you can start from.
A few important disclaimers
I am not a professional developer, this is a hobby project. This means a few things:
This project is developed in my free time, i can not gurantee continued development, this depends on my motivation and time
Things will be buggy
I relied heavily on the support of LLM Models for coding this (again not a professional developer). While i spend a lot of time on the solver, the UI is essentially “vibe coded”. I think this is espicially important to disclaim, since AI use is an (understandibly) sensitive topic in the art world.
Other people are invited to contribute to this project
How to use:
After you installed the addon you can find a new editor called Continuum Flow (where things like uv editor or geometry nodes editor are). There you can setup the node tree for your simulation. The follwing screenshot shows an example:
No problem, i would not call myself an expert either .
The simulation tool needs python and a bunch of python libraries. Since i do not want to bother people with installing that themself, i use uv. Thats essentially a tool that does that for you. You can find more information here if you want (https://docs.astral.sh/uv/getting-started/installation/).
You need to open powershell (if you are on Windows, i have not tested the addon on other os) and then simply copy the command into powershell and hit enter.
Hi thanks for the screenshot. Maybe i have time to look into this in the next couple of days. I mainly developed this for Windows. I can probably get the CPU solver to work on Mac, but the GPU is probably of limits, because the combination of python+cuda+mac is complicated.
This looks very promising, thanks for sharing! Kudos for the well done trailer. The campfire in particular looks very convincing. Will keep my eye on this and try it at a later stage…hopefully I can get it to run on Linux.
Specifically the sections on Python wheels and packaging an addon/extension. Since Blender 4.2 you can ship wheels directly inside the extension via blender_manifest.toml, and Blender handles isolated per-extension site-packages for you. That would replace the entire uv bootstrap - no separate env on disk, no PATH edits, no “Install Solver Environment” button, clean uninstall. numpy is already bundled with Blender so you can drop that dep entirely; scipy and trimesh bundle cleanly; numba needs per-platform wheels but is doable; numba-cuda is the only real complication since it still needs a host CUDA runtime regardless of how it’s installed.
Your current install procedure i would not touch with a stick. Integration is half the work, unfortunately.
Yes very likely so. I am currently rebuilding a lot of stuff. It will very likely work in newer Blender versions in the future, but i can not say right now when thats going to be.
That can be implemented at least in some form, but it is more at the end of my priority list
I am a bit unsure what you mean. Currently the domain can not be moved. (of course you can move the simulation result if you want to). I plan on implementing an option to move the domain, but that is again not the highest priority
.
