MeshGod/TRELLIS Mixamo re-rigs pose half a metre underground and compressed
(bone-space origin at hips, 0.01/100 scale sandwich). Instead of fighting
transform conventions, the renderer probes the first posed frame, measures
the actual sprite in pixels, and servo-corrects ortho scale + camera z:
unclip -> converge height -> pin feet to the ground row. Also: normalize
scales all root objects, deformed-height helper.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
1005422053