I save her as portrait_final_12.duf and close the window. Tomorrow, I’ll adjust the left eyebrow by 1%. Because perfection is a decimal, not a destination.

This is the art of the slider: Mouth corner pull — 42%. Cheek hollow — 0.17. Skin glossiness — “morning dew, but not sweat.”

She is not real. But at 4.20, with post-denoiser and tone mapping, she looks like she’s about to speak.

The progress bar fills. 90%… 95%… “Render completed — 2.3 samples per pixel.”

I never give her a line. Silence is the most human thing I can render.

Then lighting. Three-point, but with a fourth — an HDRI from a rainy Tokyo alley, 2019. The shadows stretch long and blue, like regret rendered at 4K.