OpenCV solveCubic_0

solveCubic_0 finds the real roots of a cubic equation: a·x³ + b·x² + c·x + d = 0. You hand it the coefficients, it hands you the roots. It's cv2.solveCubic wrapped as a node, and if that sounds like a strange thing to find in an image tool, you're not wrong - but it earns its place in the geometry corner of this pack.

Where this actually gets used: cubic equations show up whenever you're solving for distances or angles from geometry - computing the pose of a camera, intersecting rays with surfaces, certain fit problems. OpenCV implements solveCubic as a building block for exactly those tasks. In a ComfyUI workflow, you'd reach for it as part of a custom geometry or calibration chain, not for everyday image generation. It's a "the graph needed a math primitive" node.

The mechanism is a closed-form cubic solver (Cardano-style) that returns only the real roots. Inputs:

• coeffs - NPARRAY, the coefficients. OpenCV accepts a 3- or 4-element array. With 4 elements it's [a, b, c, d] for a·x³ + b·x² + c·x + d = 0; with 3 elements the leading coefficient is assumed to be 1, so [b, c, d] means x³ + b·x² + c·x + d = 0. Most callers pass 4.
• Optional roots - the out-parameter. Leave it unconnected; the node allocates it.

The two outputs are a tuple, and both matter:

• int - the number of real roots found. A cubic always has at least one real root, so expect 1 to 3.
• nparray - the roots themselves. If fewer than 3 real roots exist, the array is padded with non-roots (typically zeros or NaNs), which is why you check the int output first.

Here's the beginner trap, and it's straight from the pack's own troubleshooting: this nparray is not an image. The README warns that "not every return type of nparray is actually an image" - some are just 1-D arrays of floats - and if you wire the roots output into Nparrays2Image you'll get 'NoneType' object has no attribute 'shape' or garbage. The roots are numbers to be consumed by other math, not pixels to be viewed.

And the usual twin: solveCubic_1 is the identical overload - same inputs, same outputs, both call cv2.solveCubic. The auto-generator numbered them; pick one.

Install: ComfyUI Manager → search "opencv-comfyui", or git clone https://github.com/geroldmeisinger/opencv-comfyui into ComfyUI/custom_nodes, restart. Needs opencv-contrib-python plus numpy/torch (already present). Watch for the known conflict at startup - Cannot import name 'guidedFilter' from 'cv2.ximgproc' - which means two OpenCV installs are fighting; uninstall one.

Honest verdict: this is a niche node for niche pipelines, and if you don't do geometry math in ComfyUI you'll never touch it. But it's a good example of the pack's real value - giving you the complete set of OpenCV math primitives in graph form, so a workflow that needs a cubic solve doesn't have to escape to a Python script node.