Resources Numerical accuracy Calculation policy
Turns a glTF or FBX material's numbers (roughness, index of refraction, film thickness, emission strength) into the renderer's parameters and back.
Each parameter is copied or converted in 64-bit, and every number in the result is then rounded to 7 significant digits with toPrecision(7). The values describe how a surface is shaded; they are kept on the import report.
Open item. Keep the converted parameters unrounded; toPrecision(7) drops digits the source file had.
4 calculations in 1 source file. The code is read from the released app when this page loads. It is the calculation people run, not a copy kept for this page.
Tests tools/modelcraft-test/gltf.mjs
Rounds each number of a converted material.
v → the number nearest v with 7 significant digits
Number(v.toPrecision(7)) is applied to every numeric parameter and to each number of every list before the material is returned. This rounds the stored value, not a display.
const round = (v) => (Array.isArray(v) ? v.map(round) : typeof v === 'number' ? Number(v.toPrecision(7)) : v);A thin film's thickness in micrometers and a glass's Abbe number.
thickness (µm) = maximum thickness (nm) ÷ 1000; Abbe number = 20 ÷ dispersion
A division by the whole number 1000 and one by the file's dispersion, which is checked to be above 0 first.
p.thin_film_thickness = m.iridescence.thicknessMax / 1000;
p.thin_film_ior = m.iridescence.ior;
}
if (m.anisotropy && m.anisotropy.strength > 0) p.specular_roughness_anisotropy = clamp01(m.anisotropy.strength);
if (m.dispersion > 0) { p.transmission_dispersion_scale = 1; p.transmission_dispersion_abbe_number = 20 / m.dispersion; }Splits an emissive color into a color with a peak of 1 and a luminance.
peak = largest channel; color = channels ÷ peak; luminance = peak × strength
The division by the peak runs only when at least one channel is above 0, so it is guarded.
const peak = Math.max(...m.emissive);
p.emission_color = ap1(m.emissive.map((c) => c / peak));
p.emission_luminance = peak * strength;A roughness from a Blinn–Phong shininess exponent.
roughness = (2 ÷ (s + 2))^¼, clamped to [0, 1]
Two square roots of a quotient whose denominator is at least 2, because the exponent is floored at 0. The code's comment names it the usual published conversion.
const s = Math.max(0, n('ShininessExponent') ?? n('Shininess'));
p.specular_roughness = clamp01(Math.sqrt(Math.sqrt(2 / (s + 2))));Every value is a 64-bit number from input to result. Formulas are closed form or a direct solve, and every comparison uses a stated tolerance.