A genuine click inside the source canvas picks one of two locally drawn symbols, dog or cat. It appends one shape at that click coordinate. Two separate inside clicks therefore produce two stored shapes, represented by two markers. No pet photo, prompt, classifier, language model or website generation occurs. The p5 library draws fixed geometric silhouettes; the animal type is a local pseudorandom choice, not a learned interpretation of the click. The dog body is a sixty by forty canvas-pixel ellipse. The cat body is fifty-five by thirty-five. At three video pixels per source pixel, the outlines are one hundred eighty by one hundred twenty and one hundred sixty-five by one hundred five. Ear triangles use their source coordinates at the same scale. The dog reaches forty pixels above the click through its ears; the cat reaches thirty-five. This geometry explains the fixed icons, not animal anatomy or an image-model output. The source uses the smaller of viewport width and eight hundred pixels for canvas width, and the smaller of sixty percent of viewport height and five hundred for height. A thirteen-hundred-wide viewport therefore gives width eight hundred; a three-hundred-ninety-wide viewport gives three hundred ninety. Resize redraws retained shapes at their original coordinates, so a narrower canvas can clip them. Reset clears the local shape array. Download exports the actual current canvas PNG; reload starts empty. No generated website or backend model is involved.