FSR2/3 takes a Transparency & Composition mask alongside the reactive mask. We deliberately ship only reactive (explicit + auto-generated from the opaque-vs-final color diff).
Deferred with reasons, on 2026-07-10:
- T&C is not a second reactive channel. In FSR2 it applies a softer history-distrust than reactive, widens the rectification AABB, and interacts with locks — genuinely distinct, genuinely tuned behavior.
- Implementing it honestly means new tuning constants plus edits to the accumulate blend and lock path — the most protected code in the project.
- Shipping it as a channel that behaves identically to reactive would be misleading — an API that implies FSR parity while doing something else.
- The common three.js transparency case is already covered by reactive + the auto-generator (
examples/05-transparency is the acceptance demo).
Revisit when someone is authoring real T&C content and can show reactive is insufficient for it — ideally with a repro scene. That evidence is what would justify touching the accumulate tuning; without it this stays closed-by-default.
If you have that case, comment here with the scene and what reactive fails to do.
FSR2/3 takes a Transparency & Composition mask alongside the reactive mask. We deliberately ship only reactive (explicit + auto-generated from the opaque-vs-final color diff).
Deferred with reasons, on 2026-07-10:
examples/05-transparencyis the acceptance demo).Revisit when someone is authoring real T&C content and can show reactive is insufficient for it — ideally with a repro scene. That evidence is what would justify touching the accumulate tuning; without it this stays closed-by-default.
If you have that case, comment here with the scene and what reactive fails to do.