Pet Companion · Desktop Pet Builder

Problem

When turning a character image into a desktop pet, the hard parts are not just “make a window move.” The result needs clean transparency, no gray halo, consistent canvas alignment, distinct idle animations, coherent click feedback, and a reproducible packaging layout. This skill targets that concrete engineering problem: given one character image or a set of expression images, produce a desktop pet with transparent always-on-top behavior, dragging, resizing, roaming, and interaction mapping.

How It Works

• Image pipeline: process_v6.py removes watermarks, cleans white/light-gray backgrounds with flood-fill, removes edge halos, locks the largest connected component, and aligns outputs to a shared transparent canvas.
• Expression animation: nine states ship with independent MOTION_PROFILES, so each face can breathe, sway, bounce, or wiggle differently. QPropertyAnimation drives anim_rotation, anim_scale_x, and offset properties for smooth idle motion.
• Interaction and roaming: click actions map to expressions and speech bubbles, while autonomous walking/swimming uses edge bouncing, facing flips, and periodic state changes. A do-not-disturb mode hides the pet during full-screen work or text input.
• Packaging: product/ is the canonical package with pet.py, nine character_*.png files, and build scripts for macOS .app or Windows .exe flows.

Boundaries

• Best suited for relatively clean character images or teams willing to validate pixel metrics such as background residue and edge halos.
• Generating new expressions with image-to-image models can introduce eye artifacts or asymmetry, so a processed multi-expression set is safer.
• Runtime depends on PyQt6, Pillow, NumPy, and SciPy; Windows executables should be built on a Windows host.