Tools used and why

Tool	Role	Why this one
🐍 Python 3.8+	the whole program	everywhere, easy to read and change
🔢 numpy	sampling engine, balancing, reverb	sums thousands of notes in a few seconds, nothing to compile
📄 PyYAML	reads the song file	YAML is readable and easy to write by hand
🎬 ffmpeg	amp, cabinet (afir convolution), EQ, compressors, reverb, limiter, MP3, FLAC conversion	professional audio filters in C, very fast, installable on every system
🎹 SFZ	instrument format	open standard (text + WAV): free quality libraries with clear licenses
🎸 Black & Green Guitars	guitar samples (Gretsch)	sampled on every semitone, with staccato; recorded direct (DI), so the amp is chosen later
🔈 Jester's IR	guitar cabinets	impulse responses of miked Marshall 4×12s: the sound of a real cabinet, via convolution
🥁 Salamander Drumkit	drum samples	a real acoustic kit, many dynamics and round robins: no "machine-gun effect"

Why not a General MIDI soundfont with fluidsynth? That was the first version: quick to write, but GM guitars sound fake. Here every instrument is a dedicated multisampled library, and the guitar goes through a simulated amp as in a studio.

Why not use an external sampler directly (sfizz, LinuxSampler)? They are not available evenly on every system. The built-in numpy engine reads only the subset of SFZ it needs and runs wherever Python runs.