OpenCV haveOpenVX_0

haveOpenVX_0 is the purest example of "this pack wraps every top-level OpenCV function, no matter how useless that is in a node graph". It takes no inputs, returns one boolean, and that boolean is going to be False for essentially everyone reading this. Which is a fine reason to understand it so you can ignore it with confidence.

What OpenVX even is

OpenVX is Khronos' cross-platform vision acceleration API - a specification for running computer vision kernels (blur, edges, optical flow) on specialized hardware like DSPs and GPUs. It's aimed at embedded and mobile SoCs. OpenCV can optionally be built with OpenVX support so some operations get accelerated underneath.

cv2.haveOpenVX() reports whether the OpenCV you imported was compiled with that support enabled.

Why it says False

Here's the thing: the OpenCV wheels you install with pip install opencv-contrib-python are generic desktop builds. OpenVX support is almost always disabled in them - it requires a hardware target and a matching OpenVX implementation at build time, which is exactly the kind of platform-specific thing generic wheels don't ship. So unless you compiled OpenCV yourself against an OpenVX backend (and if you had, you wouldn't be here), this returns False.

There is nothing you can do about it from ComfyUI. You can't toggle it, you can't install it, and there's no node to attach it to. It's purely diagnostic information about your build.

Why the node exists

The opencv-comfyui pack auto-generates one node per top-level OpenCV function. It filters out functions whose types it can't represent - classes, callables, Sequence returns - but a zero-argument function returning a boolean survives the filter easily. So you get haveOpenVX_0, a node whose output is a fixed False on standard installs. It's a fossil of the generator, not a feature anyone designed.

Install

It's part of the pack, so:

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

or "OpenCV" via ComfyUI Manager, plus pip install opencv-contrib-python.

Should you use it?

No. If you're building a workflow that genuinely needs to know whether your OpenCV has OpenVX acceleration - which would only matter on embedded hardware - this is your node, and True would be meaningful. On a normal desktop ComfyUI install, wire it to a text/display node once to satisfy your curiosity, then remove it. It's the pack's equivalent of a "you are here" sticker on a map of the ocean.

This is also the pattern to recognize across the whole pack: with 635 auto-generated nodes, a fraction are things you'll never touch. haveOpenVX_0 is the poster child.