nx_phototwin.nx

nx_phototwin.nx -- THE PHOTOGRAPH-TO-SUBSTRATE TWIN (2026-08-24, /compare/phototwin PT2 PT3 PT7). license_tier: ORIGINAL No hw writes (Rule 26). nx_phototwin report <image> <image.icc> <substrate.icc> [conf] gamut-coverage report of a photograph nx_phototwin proof <image> <image.icc> <substrate.icc> <outdir> [conf] twin.png (chroma-clipped) + heat.png (dE00 map) + proof.txt nx_phototwin wash <ledger.tsv> <sku> [conf] wash-fastness ledger verdict for one SKU nx_phototwin bench <ledger.tsv> [conf] the swatch bench over EVERY sku, refusing partial coverage EXIT: 0 OK | 1 REFUSED (tolerance exceeded, wash failed) | 2 usage | 3 UNOBSERVABLE (undecodable image, unreadable profile or conf, no ledger rows) -- the third state is never a pass and never a failure of the subject. TWO PROFILES, TWO ROLES. Pixels are interpreted through the IMAGE profile (RGB -> TRC -> matrix -> XYZ(D50) -> CIELAB, one ruler, nx_colorsci_lib) and tested against the SUBSTRATE profile's gamut -- EXACTLY for a matrix/TRC profile (linear RGB through its inverse in [0,1]; the cube is the gamut). Out-of-gamut colours are chroma-clipped at constant L and hue (the twin), then rendered back through the image profile so the soft proof displays in the image's own space. dE00 between photograph and twin is the proof: mean and max against knowledge/colorsci.conf (printing.org bar). MEASURED 2026-08-24 before this shape: one profile for both roles made the narrow-substrate control VACUOUS (the image's codes were read as substrate codes and were inside by construction), and a segment-maxima descriptor read 18 percent of an sRGB photograph outside sRGB -- both retired here. The descriptor stays in the lib for MEASURED swatch gamuts (PT7), which have no matrix. The mapping is MEMOISED PER EXACT 24-BIT COLOUR (a photograph has ~1e5-1e6 distinct colours, never 4e6); the unique-colour count is printed so the cost is on the record. SUBSTRATE HONESTY: until a swatch bench measures the pigment-on-Supima profile, a proof whose substrate is the image profile itself is a positive CONTROL and says so in proof.txt. WASH LEDGER (knowledge/phototwin_wash.tsv): rows swatch sku blank line cycle L a b, Lab measured before (cycle 0) and after declared ISO 105-C06 cycles; dE00 per cycle against cycle 0; a SKU with no rows is UNMEASURED (exit 3).