BD AutoRig (UniRig) — TBD

I'll save you the click: don't use this node. BD AutoRig (UniRig) - TBD is a placeholder, registered so the pack's AutoRig category has a home for an integration that isn't built yet. Run it and it raises NotImplementedError - it takes a mesh input, promises an fbx_path output, and delivers neither. There's no hidden functionality, no workaround toggle.

What it's supposed to become: UniRig (VAST-AI Research, MIT-licensed), a SIGGRAPH 2025 auto-rigger. The problem, per the pack's own source, is that UniRig's dependencies - spconv, flash_attn, torch_scatter, torch_cluster - are a compiled-CUDA nightmare that won't auto-install cleanly in most ComfyUI environments. That's the same class of install fragility the KB's 3D generation doc flags as the #1 barrier in this corner of the ecosystem: "the install is the real barrier, not the VRAM." Rather than ship a node that fails at import time and takes the whole pack down with it, the author registered a stub.

What to use instead

BD_AutoRigMIA - the pack's Make-It-Animatable auto-rigger. It's vendored in (no external install), is faster (well under a second vs. seconds), gives better humanoid quality, and has no dependencies beyond weights that auto-download on first use. The pack's own notes call it the right choice for humanoid characters, full stop.

The UniRig work is tracked as "ARTS-37 follow-up. PRs welcome." If you need UniRig's specific output and can stomach a manual build of spconv + flash_attn against your exact torch/CUDA, the upstream project is VAST-AI-Research/UniRig - but if you're reading this to fix a workflow problem today, you want BD_AutoRigMIA → BD_AutoRigUEFN → done.

Install (for completeness)

The node ships with the pack, so if you've installed ComfyUI-BrainDead it's already there:

cd ComfyUI/custom_nodes
git clone https://github.com/BizaNator/ComfyUI-BrainDead
cd ComfyUI-BrainDead
pip install -r requirements.txt

It won't hurt anything sitting in your node list - it's inert. The only reason to care about this node at all is so you don't burn ten minutes wondering why "AutoRig (UniRig)" errors the moment you connect it. It's a roadmap marker, not a tool. Check the GitHub repo's issue tracker before you assume it's been implemented since this article was written.