Converting a colour does not make it fit.

A colour can fit a larger space but require negative or above-one components in a smaller space. After correct linear conversion, such values are not necessarily errors: the exact colour lies outside the chosen bounded RGB cube.

Identify the processing stage before limiting it. Extended values may be useful in scene-referred work; final output has defined constraints. Requiring 0…1 at every node can destroy data prematurely.

With matched white points, a linear XYZ conversion changes RGB coordinates for the same stimulus. It does not choose a replacement for an unreproducible colour. Different whites add an adaptation decision, which is not itself complete gamut mapping.

The lab’s Rec.2020, Rec.709 and P3 share D65, so no CAT is added. Unbounded target RGB and XYZ are computed first, then two mapping branches. This separates exact coordinate conversion from intentional colour change.

Per-channel clipping replaces negatives with zero and values above one with one. Unequal channel contributions to Y mean that saturation is not the only property affected: hue, light and colour differences can change too.

Clipping is a useful diagnostic for what a crude bounded conversion loses. Passing a 0…1 check does not prove appearance preservation. Inspect numerical changes and real footage.

Compression may begin inside the hard boundary to avoid an abrupt transition. Some in-gamut saturation is traded for smoother out-of-gamut behaviour. Understand the threshold and which values stay unchanged.

This lab moves colour along a ray from neutral [Y,Y,Y] in target linear RGB. Distances below the threshold are untouched; larger distances compress smoothly toward the boundary. The simple method preserves Y for a valid neutral but does not claim an optimal visual compromise.

The same “move toward grey” means different paths in different coordinate systems. A straight line in linear RGB or xy is not guaranteed to have constant perceived hue. The perceptual result depends on the model and viewing conditions.

The lab therefore claims only its verifiable linear-Y invariant, not perceptual accuracy. ACES 2 performs its corresponding processing in perceptual polar coordinates with target constraints, a substantially more complete model.

An xy diagram discards the absolute XYZ scale and shows chromaticity. It helps inspect primaries and shifts, but cannot alone establish whether a bright saturated stimulus fits a real screen. Light and channel constraints both matter.

Changing sample level keeps source chromaticity while changing distances to RGB-cube faces. The mapping can respond differently. Browser swatches are always limited to sRGB; for P3 this is an extra preview limit, not evidence that the calculation lacks P3 colours.

Wider primaries do not mean every pixel uses the full gamut. Many colours in a Rec.2020 file already fit Rec.709 and need not receive the same compression as extremes. Inspect content, not just the tag.

Compare two destinations for the same sample. The matrix and target cube change while the source stimulus remains fixed. Extremes often reveal larger differences; neutral is necessary but insufficient. A real screen may not fully cover the named target either.

Scene-referred compression can tame difficult saturated values and improve downstream stability. Its purpose is not necessarily to place every value immediately inside a final display RGB 0…1 range.

Output mapping accounts for a concrete destination together with tone rendering. The two stages are not interchangeable because both mention gamut. Working-signal conditioning alone does not promise a finished Rec.709 image.

ACES 1.3 Reference Gamut Compression is a defined scene-referred conditioning algorithm with a trust region. Its specification does not require every output to lie inside AP1. That differs from a blanket promise to remove all out-of-gamut values.

ACES 2 Output Transform gamut compression operates within target-aware output rendering. Do not transfer one stage’s name or properties to the other. The lab’s ray model implements neither; it illustrates only its stated invariants.

Before choosing gamut fitting, confirm the source primaries and transfer. An input-assignment error can look like excessive saturation; compensating compression merely hides the cause. Correct interpretation first.

Then identify processing already owned by the project, ACES or a CST. Avoid habitual extra limiters. Inspect neon, skin and gradients, including their interaction with tone compression. Available controls depend on Resolve mode and version.

Check meaningful colour differences, boundary smoothness, neutrality, luminance and hues in real footage. A method may fit every pixel in the RGB cube while producing poor gradients or awkward highlight colours.

Diagrams and numbers help locate changes. Here, sRGB swatches may look identical after preview clipping even when their calculated XYZ differ. Do not infer numerical or perceptual equality from browser colour alone.