The 3-mode body read as a few isolated peaks, not a resonant instrument body. A real corpus is a 3D resonant volume — a dense field of eigenmodes. Replace it with a 14-mode body bank modeling the violin-corpus signature: A0 air resonance, the low wood plate modes (B1-/B1+ ~460-530), mid-wood modes, and the bridge-hill cluster ~2.2-3kHz — anchored on bodyf[0] for size + bodyg[0] for level, so cello/bass inherit the same modal SHAPE an octave+ lower. This is the efficient 'volumetric' body: the corpus's 3D eigenmodes baked into parallel resonators (a literal waveguide mesh is too heavy for polyphonic real-time). Output-only — pitch/loop untouched. Bonus: the coherent modal spectrum tightened pitch detection — all 5 bowed strings (Violin/Viola/Cello/Contrabass/Tremolo) now PASS clean. @jeffrey confirmed it reads as a fuller resonant body. New GMVoice rbody_* fields (up to 24 modes).