port-wmp-viz

Extract the visualization DLL from its installer with cabextract, decompile it using Ghidra headless with the DumpDecompiled script, analyze the render algorithms, then write a .milk preset mapping the logic to per-frame equations and HLSL composite shaders.

Most WMP viz installers are MS CAB self-extracting archives, so cabextract works directly. If the CAB is embedded at a non-zero offset, search the binary for the MSCF magic bytes, slice from that offset, and run cabextract on the extracted slice.

Black screens usually come from missing MILKDROP_PRESET_VERSION and PSVERSION headers before [preset00], which makes ProjectM default to MilkDrop 1.x without shader support. Also check for fDecay=0.0, gaps in comp_ line numbering, or // comments inside backtick-continued shader lines.

Per-pixel and per-frame EEL lack floor, ceil, step, smoothstep, lerp, if/else blocks, and loops. Use workarounds like min/max for clamping, ternary if(cond, a, b), and manual smoothstep expansion; reserve floor and tex2D for HLSL shaders.

Yes, use visual approximation instead. Analyze original screenshots with PIL and numpy for color palettes, symmetry, and spatial distribution, then recreate the effect with math-simulated patterns like spirals and radial gradients in a composite shader.

You need Ghidra 12+ with Java 21 for decompilation, cabextract for installer extraction, Python 3 for the syntax validator, a C++ compiler for the parser test harness, and a ProjectM build for visual testing.