The look profile

A .vxlook is one asset holding the artistic values a whole project shares: the exposure it is pinned to, the range its meter may wander, its grade, its fog, its vignette. It is the bottom of the post-process volume stack — Unity's Volume Profile by way of Vixen's own overlay model — and its payload is PostProcessSettings, the same per-parameter opinion struct a volume authors:

Every field is optional — doc 32's "says nothing / has an opinion" distinction — so a look that only pins the exposure says nothing about anything else, and null is never 0.

⚠ fogColour is a radiance in cd/m², and it is the one field here a colour picker will get wrong. Everything else on this list is a ratio or a stop; this one is a luminance the frame's own scene colour is lerped toward, and a lit frame's sky averages thousands of it. A value near one is not a subtler fog — it is a fade to black that is too faint to see, which is exactly how it went unnoticed in three separate passes. Prefer saying nothing: a !Fog node with no opinion laid over it takes the sky's mean radiance and the sun's illuminance from the scene, which is the number this field is trying to guess. See FogRenderer.Colour.

The precedence is fixed, four layers, never ambiguous:

The look loses to anything any volume says, per parameter, whatever the volume's priority — the precedence is by kind of layer, not by number.

The Standard Frame deliberately emits neutral artistic values, because the audit behind doc 39 showed what art living inside pipeline structure costs: sample 13's dusk works only because ev100, the meter's clamps and the fog colour all agree with the sky, and those agreements were scattered across eleven hundred lines. They belong together, in one named artifact the editor can edit live — and because the look reaches the frame through the volume fold at run time rather than through the expansion, editing it relights the same document with nothing rebuilt.

The boundary with RenderQuality.vxpreset is one rule: look changes the intent, quality changes only the fidelity and cost of the same intent. Bloom threshold and knee are look; bloom pyramid levels are quality. DoF aperture and focus distance are the camera's and live on neither — the look's only word on defocus is the volume vocabulary's maximumDefocus ceiling.

⚠ Exposure is the one absolute in the stack. A volume says exposureCompensation — an offset that composes — but the look may pin ev100 itself, because "this game's dusk is EV 12.5" is the base those offsets compose over. The pin blends in stops, not linear exposure, and only reaches frames whose exposure is fixed: a metered frame reads the meter's buffer, and the look constrains that through meterMinimumEv/meterMaximumEv instead. The clamps are authored as EVs and converted where consumed — the meter's own knobs are reciprocal linear multipliers, and that inversion stays out of every document.

Two ways in, and the document's wins:

The seam underneath is small and stated: the !StandardFrame transformer deposits the document's inline look on CompositorBuilder.Look (an output of the build — the look is never baked into the emitted nodes, and the expansion with and without one is snapshot-identical); after Load, the host hands whichever look won to PostProcessVolumeSystem.Look, where the fold lays it down first at full weight. A game holding the asset already may assign graphics.Volumes.Look directly.

⚠ A look opinion reaches nodes whose own knob is off in the variant sense. The Standard Frame's tonemap has no grade compiled in; a look's grading: turns the permutation on and blends from ColorGrading.Neutral — never from a zeroed struct, whose gain of zero is a black frame.

Asking why it looks like this — the resolved stack, per camera, bottom layer first:

A scene overriding one thing. The look pins the whole project's fog; one level wants it thicker. An unbound volume with fogDensity: 0.05 says exactly that and nothing else — every other parameter stays the look's, which is the entire point of per-parameter opinions.

Switches flip at half weight. fogHeightFalloff and fogSunScattering are booleans, and a weighted fold cannot crossfade one — so a BlendedToggle takes the heavier side, flipping at the blend's midpoint. The look always contributes at weight 1, so this only shows where a bounded volume claims a switch and fades.

• Making a room look different — the volume system the look is a layer of, and the opinion model both share.
• The Standard Frame — the neutral frame the look dresses, and the look: knob.
• Scaling the frame with render quality — the other named asset, and the intent/fidelity boundary between the two.
• docs/plan/39-standard-frame-and-render-presets.md § The look profile — the design and the four-layer precedence.