OpenCV getBuildInformation_0

Drop getBuildInformation_0 into a workflow, run it, and read the output. What you get is a wall of text - the full build report of the OpenCV that's installed in your ComfyUI environment. Version, CPU optimization flags, which modules are compiled in (contrib modules? CUDA support? OpenCL?), operating system, compiler. It's the same string cv2.getBuildInformation() returns in a Python shell, exposed as a node.

Why would you want that? Because this pack's nodes wrap whatever opencv-contrib-python is installed, and the pack is full of functions that only exist if your OpenCV build has the right modules. The README's known-conflict error - Cannot import name 'guidedFilter' from 'cv2.ximgproc' - is the poster child: someone's OpenCV is missing or conflicting, and this node is exactly the tool for confirming what you're actually running. Before you blame a node for failing, check whether the function is even compiled in. This node tells you, instantly, without leaving ComfyUI.

It's also the answer to a question beginners hit early: "why does this pack work for one person and not another?" Because OpenCV is a build matrix. Pip wheels differ by platform, CUDA, and which contrib modules are present. The 635 nodes in this pack only make sense against the specific build behind them.

Inputs and output

No inputs. It's a pure "read the environment" node. Output is string, the build information report.

That's the whole interface. There is nothing to configure, which makes it the least error-prone node in the pack - and, fittingly, one of the least "node-like" things you'll ever put in a graph. Some people wire it to a text preview just to inspect it; others run it once when troubleshooting and remove it.

Installing

Part of geroldmeisinger/opencv-comfyui. ComfyUI Manager: search "opencv-comfyui". Or:

cd ComfyUI/custom_nodes
git clone https://github.com/geroldmeisinger/opencv-comfyui

Restart ComfyUI. The pack's requirements.txt installs opencv-contrib-python (plus numpy and torch) - and yes, the -contrib part matters, because the contrib modules are where many of the fancier functions live.

Common issues

Honestly? This one can't fail in the usual ways - no images, no batches, no literals. The only failure mode is an environment without OpenCV at all, which would break the entire pack at load time, not just this node.

Its real value is diagnostic. When another node in the pack throws a missing-function error, drop this one in, read the report, and you'll know whether your build has the module, whether CUDA is compiled in, and whether you're on the version you think you are. It's the pack's own "is my OpenCV healthy?" node, and it's the first thing I'd reach for when the dragons start breathing.