What actually shipped
The project now includes an official add-on for Blender that connects to an external solver backend, rather than replacing Blender’s built-in simulation stack. The add-on acts as a client and runs anywhere Blender runs, including macOS, while the solver engine runs separately. And if you understand higher physics, here is an explainer. Fort the rest of us:
ZOZO’s Contact Solver codebase lives on GitHub, and the solver is a contact solver for physics-based simulations involving shells, solids, and rods. The project uses the Apache License 2.0, which you may use, modify, and redistribute the code in commercial products, including proprietary software, as long as you preserve the required notices.
A release tagged “Blender Add-on 1.0.6” has support for armature-driven deformable colliders and bug fixes related to paths containing empty spaces. That is a very production-shaped patch note, because file paths love whitespace even when pipelines do not.
The headline features, minus the marketing glitter
The add-on is exposing finite element deformable and penetration-free contacts, plus strictly strain-limited cloth intended not to stretch like rubber. That last bit aims at the old pain point where cltoh behaves like chewing gum under stress.
The tool targets cloud-deployed GPUs, with examples such as vast.ai and AWS, while still allowing local runs if you have a powerful NVIDIA GPU. It also states the workflow can run locally or remotely on Windows and Linux, with macOS supported via a remote solver engine because CUDA does not run there. Why would it.
The core technology is backed by a peer-reviewed publication in Transactions on Graphics for SIGGRAPH Asia 2024.
Hardware and deployment reality
The installation makes the split explicit: the add-on is just a client, but the solver backend requires an NVIDIA GPU with CUDA 12.x. That makes the tool’s core constraint very clear: your workstation can be lightweight, but your backend needs serious GPU muscle. There are multiple ways to connect to a backend, including running on the same machine, using an SSH-reachable Linux host, using a Docker container, or using a Windows workstation. In other words, it is built for the common studio pattern where artists drive a tool from a DCC while compute happens elsewhere.
The repository also includes a section on deploying on cloud services and includes example hourly costs for specific setups, such as approximately €0.76 per hour for a Scaleway setup and around $1 per hour for a described Amazon instance configuration. Those numbers relate to the referenced cloud configurations, not a license fee for the tool itself. And not the pricing when you did saomething a little bit wrong and are accumulating AWS-bills like a torado in a bank. You know what I am talking about.
