I run a 3D ghost-mesh scan with RealityCapture to capture the facial structure beneath the noise, anchoring the brow, cheek planes, mustache, and hat brim in geometry instead of guesswork. From that scaffold, I use AI as a restrained assist to stabilize features and rebuild continuity without letting it overwrite the source.

ll add skin and surface color through an inclusive, evidence-led parameter set that treats complexion as a product of light, environment, material, and imaging chemistry, not a stereotype. I’ll model illumination first by estimating time-of-day color temperature, shadow softness, and ambient bounce, then I’ll tune regional atmosphere through altitude, dust, humidity, and haze so the face reads like it exists in that landscape’s physics.

I’ll calibrate skin using melanin-forward values, undertone balance, and specular response, cross-checking against period-appropriate film stock behavior and the scan’s tonal range so the highlights don’t bleach and the midtones don’t collapse.

I’ll keep the AI pass constrained to continuity and denoising, then I’ll finalize in Procreate with measured shifts in warmth, pore texture, and edge falloff that match the mesh and the original grain, while documenting uncertainty where the archive doesn’t give enough information to claim precision.