TL;DR: With Substrate enabled and Substrate GBuffer Format = Blendable GBuffer, a hand-built Slab BSDF loses specular color (hue) when F0 is dark (below the typical conductor range). The same input values through the legacy MaterialAttributes path keep their colored highlight. This looks like a lossy repacking step when writing F0 into the legacy-shaped GBuffer, not a tonemapping artifact (verified below).
Repro (UE 5.8, Substrate on, Blendable GBuffer):
I have two materials with identical inputs — Constant3 (0.092, 0.0829, 0.0106) (dark gold), Metallic = 1, Roughness = 0:
- Legacy path: MakeMaterialAttributes → root node
- Substrate path: same constant → Substrate Metalness-To-DiffuseAlbedo-F0 → Slab BSDF → Front Material
Result: Material 1 renders a gold-tinted highlight. Material 2 renders a pure white highlight, and the surface reads less metallic overall (diffuse contribution appears to increase).
[Image 1: legacy path, golden highlight]
[Image 2: Slab path,white highlight]
Additional data points:
- With a bright base color, e.g.
(0.63, 0.57, 0.07), both paths render identically (gold highlight). The divergence only appears as F0 luminance drops — consistent with F0 falling out of the conductor range (docs state conductors ≈ 0.5-1.0, dielectrics ≈ 0.02-0.06) and being repacked as a dielectric: chroma collapsed to a scalar specular, residual energy pushed to diffuse. - Not tonemapper clipping: reducing the light from 4 lux to 0.4 lux (highlight far from clipping) — the Slab-path highlight is still white.
- Verified the conversion node itself is correct: routing its F0 output into Diffuse Albedo displays the expected dark gold, so the chroma is lost after the material graph, at GBuffer write time.
- The Substrate panel confirms the material runs as Blendable GBuffer (“no Substrate tree – passing legacy GBuffer data” topology on the legacy path; F90 listed under Disabled features / unsupported on Slab materials).
- There is also an in-between status. when the color is (0.314, 0.283, 0.036):
Questions:
- Is this expected behavior for Blendable GBuffer (i.e., F0 outside the conductor range cannot round-trip through the legacy GBuffer layout)?
- If so, could this be documented? The current docs describe Blendable as “feature-limited” but don’t mention that dark/unusual F0 chromaticity is silently discarded — which also affects any custom colored-specular work (iridescence via F0, etc.).
- Is the only fix Adaptive GBuffer, or is there a per-material workaround?
Engine version: 5.8.