Illustrative: streak texture, temperature scale, exposure, bloom; probe marker not to scale.
What is calculated
Spacetime: Kerr metric, G = c = M = 1, spin 0 ≤ a ≤ 0.99. All radii are Boyer–Lindquist r in GM/c².
Light paths: for each pixel one null geodesic is integrated backwards from the camera (Mino-time form of the exact equations, Runge–Kutta–Nyström 4, ≤ 64 steps). The shadow, the far side of the disk appearing above and below the hole, and the n = 1, 2 photon ring all come out of this integration. Higher-order images (n ≥ 3) are thinner than a pixel and are dropped.
Disk inner edge: the prograde ISCO (Bardeen, Press & Teukolsky 1972): 6 at a = 0, 1.45 at a = 0.99.
Frequency shift: g = νobs/νemit for gas on prograde circular geodesics, evaluated with each ray’s conserved angular momentum. It contains the Doppler shift, gravitational redshift and frame dragging together.
Emitted flux: Novikov–Thorne / Page–Thorne thin-disk profile with zero torque at the ISCO.
Probe clock: prograde circular geodesic, dτ/dt = √(1 − 3/r + 2a/r3/2) / (1 + a/r3/2), accumulated against Boyer–Lindquist time t.
Approximations
Disk: infinitely thin, in the equatorial plane, opaque except for a wispy outer rim; no disk thickness, no self-irradiation, no gas inside the ISCO.
Colour and brightness: each patch is drawn with the chromaticity of a blackbody at g·T and a bolometric intensity ∝ g⁴ × flux. A real camera band would dim the receding side even more. Limb darkening is ignored.
Frozen light: each frame traces light through a disk frozen at one instant; time delays between the direct and lensed images are ignored.
Camera: at rest in the locally non-rotating frame at a finite distance; no aberration from orbital motion.
While moving: geodesics are traced on a coarser grid and interpolated, then refined to one ray per pixel when the view rests. Software renderers also lower the canvas resolution and shade sky-only areas with a reduced copy of the same shader.
Illustrative only
Streaks are procedural noise advected at the local orbital rate — not simulated turbulence.
The temperature scale (peak ≈ 5500 K at a = 0) is chosen for the eye and its rise with spin is compressed (peak T ∝ F1/8 rather than F1/4); a real disk at this mass would peak in the far UV. Auto-exposure cancels about three quarters (in stops) of the brightening with spin and inclination, so a hotter, deeper disk looks only moderately brighter; bloom, vignette and grain are cinematic.
Star field is synthetic; stars are stretched by lensing but point-source magnification is not modelled.
The probe is a small 3-D model drawn where the geodesics place its direct image (its fainter lensed ghost is a glow in the ring): enormously out of scale, not lensed in shape and not Doppler-shifted. Its ring pulses once per second of probe proper time. Mass is set to ≈ 5×10⁴ M☉ so that 1 s on the far clock equals 4 GM/c³ and orbits play in real time.