A colour palette is calibrated to its shading model, not to the screen

Fact

Colours chosen against flat fills are final values: what is written is what is seen. The moment the same colours are fed through a lit renderer they become albedo, and every stage multiplies them down — ambient term, N·L, distance attenuation, fog. A mid-dark #2c2b3a that read as "moody slate" flat can leave the screen at a few percent brightness once shaded.

The loss is multiplicative and compounds, so the darkest entries in a palette disappear entirely while the lightest merely dim. The palette does not scale uniformly; it collapses from one end.

Why it matters

Changing renderer is normally treated as a rendering task, with the art assumed portable. It is not, and the result looks like a lighting bug rather than a colour one — which sends the work into the light rig, where turning everything up only blows out whatever was already bright.

The same trap catches a flat design system moved under a translucent or blended surface, and any palette moved from sRGB fills into a linear-space pipeline.

How to apply

• Re-pick albedo against the new pipeline rather than scaling the old values. Expect to lift the dark end substantially and the light end barely.
• Sanity-check the worst case first: darkest material, minimum ambient, maximum fog distance. If that reads, the rest will.
• Keep the two palettes as separate data if both renderers survive. One shared table tuned for neither is worse than two tuned tables.