nx_tissuefield_lib.nx

nx_tissuefield_lib.nx -- THE SKULL-BOUND TISSUE-DEPTH SKIN SURFACE (aesthetictwin AT46, 2026-09-18). Operator 2026-09-18: "we were trying to render a skull then every layer up ... controllable every layer up generative capabilities". Today's faces are blended SDF blobs on a frame; nx_faceanat never reads the skull it was meant to sit on (`nxmesh` count 0 in that file). This library builds the FIRST layer ON the generated skull: the skin is an OFFSET SURFACE of the skull mesh, offset along the skull's own vertex normals by a tissue-depth FIELD interpolated from the forensic tissue-depth table at anatomical landmarks that are LOCATED ON THE SKULL ITSELF (midline profile extrema, support directions, canon ratios of the bony span), plus controllable fat compartments and named patches for the features that have no bone behind them (globes and lids over the orbits, the nose over the empty aperture, the vermilion band). EVERY NUMBER IS DATA OR IS DERIVED FROM THE RESOURCE IT DESCRIBES: depths, compartment magnitudes and construction ratios come from knowledge/tissue_depth.conf (each row names its source; rows the research lane has not pinned say PENDING-REF), the interpolation kernel width is the median nearest-landmark spacing of the located landmarks, grid steps are the mesh's own mean edge length, the units are the skull's own (NXMSH2 float32 millimetres read as tenths of a millimetre, so one tissue-depth unit mm10 IS one model unit u10), and the camera distance is derived from the head height and a declared frame fraction. Nothing here reads nx_faceanat, nx_sdfrender or nx_skullsdf as code: the skull is consumed as the MESH those organs emit, and the depth values nx_faceanat typed inline were carried into the conf with their source named. FRAME: the skull mesh is anterior +z, superior +y, the subject's left +x (nx_meshview front view = camera on +z). The rasteriser (nx_trimesh) looks down +z with the camera at -camz, so every vertex is pushed through a rotation of half a turn about y (x -> -x, z -> -z): a proper rotation, never a mirror, so the subject's right eye lands on the image left as in a photograph (the photo ruler's side convention). LIB, no main. Composes nx_nxmesh_lib (the ONE NXMSH2 layout), nx_vecmath (isqrt, float32 codec), nx_trimesh (the rasteriser), nx_png (the encoder), nx_itoa_lib (zero-alloc decimal). license_tier: ORIGINAL No hw writes (Rule 26).