How every work was measured

Every one of the 296 works in this catalogue has been measured frame by frame, directly from its master file. Not sampled, not estimated: every frame of every work. This page says how it was done, what each number means, and — the part that matters most — what the measurement cannot tell you.

What was measured

How

Each master is decoded once, from beginning to end, in a single sequential pass. Every frame is reduced to a small greyscale and colour representation and compared with the one before it. Nothing is skipped, so no event is too short to be seen: a change lasting a single frame — one twenty-fifth of a second — is still counted.

A second complete pass, in September 2026, added colour over time: the share of the frame each colour covers, step by step — every second, or in sixty steps for the shortest loops — so that a colour that stays can be told from one that comes and goes. Everything measured in the first pass came out identical, except in four works whose masters the artist has since remade, which were measured again and carry their new values.

A third pass, in October 2026, looked again at colour alone, at half the native resolution of each master, one frame per second and never fewer than sixty frames. At the small size of the first passes a thin coloured line melts into what surrounds it; at this size it does not. It gives each work its colour in the image, below.

Every threshold is relative to the work itself, never a fixed number. The catalogue spans a factor of 76 in brightness between its darkest and lightest work; a fixed threshold would be blind in half of it.

The measurements

Each work carries these as structured data. The value is the fact; the range beside it is the whole catalogue, so any number can be placed without needing an adjective.

Rhythm of change

How much the image changes per second. Not speed of motion: amount of visual change.
0.001 (El Centro (8K))  →  3.089 (La Cruz (Square))

Rhythm contrast

Whether the work breathes or never rests. A work can be slow on average and still have bursts.
1.02 (El Fresonero (8K))  →  6.28 (Ego (4K))

Composition movement

How much of the change survives blurring. Low means elements are anchored and move internally; high means the whole composition shifts.
0.062 (Vórtice en la Tierra (4K))  →  0.863 (Ego (4K))

Luminance

Mean light emitted across the whole work.
0.013 (Samsara (4K))  →  0.984 (The Dance of the Balance (4K))

Peak brightness

The 99th percentile of luminance — how bright its brightest hundredth gets.
0.129 (Samsara (4K))  →  1 (Him (8K))

Dynamic range

Distance between the darkest and brightest parts of the same frame.
0.098 (Samsara (4K))  →  1 (Lourdes Unknown (4K))

True black

Percentage of pixels at absolute black, averaged over every frame.
0 (Bésame Mini (HD))  →  93.75 (Black Maktub (4K))

Colour travel

How far the palette moves across the work.
0 (Un Mensaje Para García (4K))  →  0.752 (Julia's Journey (4K))

Flat ground

How much of the frame is one near-uniform colour.
0.092 (Space Flowers (8K))  →  0.943 (Black Maktub (4K))

Empty space

How much of the frame carries no detail at all.
0 (Superdots (8K))  →  0.886 (Tidal Love (4K))

Element scale

How much detail survives blurring. Low means the elements are tiny.
0.015 (Superdots (8K))  →  0.889 (Re (4K))

Edge hardness

Graphic and cut-out, or soft and diffuse.
0.059 (Arco Iris (4K))  →  1.218 (Wall Of Strength (8K))

Radial symmetry

How closely the frame matches its own 180° rotation.
0.158 (Duality Wall One (8K))  →  0.969 (Lourdes Unknown (4K))

Colour over time

An average colour hides time: a work that is blue from beginning to end and one where blue appears for a few seconds can share the same percentage. These measurements keep the time in.

Colour over time

For each colour family, the share of the loop in which it covers at least 5% of the frame, and how much of the frame it covers while present; a family is listed only if it is present in at least 5% of the loop. 194 works have at least one family present for the whole loop; in 65, every family comes and goes; 37 carry no colour family above 5%; in 9 of them the colour comes only in fine gestures, which the colour in the image counts.

Colour families at once

How many families share the frame at the same moment, against how many appear across the loop. Wall Of Consciousness (8K) holds 7 of its 7 at once; Días (Versión Sin Sombras) (HD) shows 7, 1.1 at a time — one after another.

Colour in the image

Whether the image carries colour the eye can see, looked at more closely: at half the master's resolution, one frame per second. 28 works are monochrome (black, white and grey), 4 of them with only the signature in colour and 3 with only a faint tint; 11 are close to monochrome, with a little colour, and say how much; 257 are in colour.

Near-black passages

Stretches where the screen is almost dark: mean light below 0.03 on a 0–1 scale for at least half a second. 17 works have them; two are near-black throughout.

Colours are read in families: blue counts azure, navy, indigo, cyan and teal; green, emerald, chartreuse and teal; red, crimson; yellow, gold and amber; purple, violet; pink, magenta; orange, rust. Teal, a blue-green, counts in both.

Verified against the masters

These are not claims about the work. They are checks run against the files themselves.

Loop closure is measured by comparing the last frame with the first, against how much the work changes between neighbouring frames at that same moment. The one work that does not close does so by the artist's decision, and says so in its own data. The single anomalous frame is likewise published, at its exact second.

What this cannot tell you

A measurement that only reports what it found is worth less than one that also says where it stops.

It measures light and change, not meaning. These numbers can tell you a work is dark, still and anchored. They cannot tell you it is good. No number here is a judgement about the art.

Composition movement is not distance travelled. It measures how much of the change survives blurring — which mixes together how far something moves, how large it is, and how much it contrasts with its ground. A big, high-contrast shape reads higher than a small, low-contrast one covering the same distance.

The photosensitivity indicator is not a certification. Recognised standards are measured in cd/m² on the actual screen. This looks at the master, and reports only that no flashing pattern was detected in it. Separately, two works carry an abrupt light event: a single-frame jump in the overall light, at least as large as the work's own average light, after which the work stays changed — Creador (HD) at 3:56 and Isabel (4K) at 3:41. It is not a flashing pattern, so it does not change the indicator; it is published in each work's data for anyone choosing for a sensitive audience.

Colour names are a convenience. The hexadecimal value is the data. Names are grouped into bands for reading, and bands are always arguable.

Colour families are a reading choice. Where blue ends and green begins is a decision, not a measurement; it is stated above so it can be disagreed with. The 5% threshold means a thin line of colour can matter to the eye without being counted as a family; the colour in the image, measured more finely, is there to catch it. And where grey ends and colour begins is a reading choice too: a pixel counts as colour from a chroma of 12, and the three works with only a faint tint are listed as monochrome by the artist's decision.

It measures the master, not your screen. What a work does in a room depends on the display, the light and the distance. The measurement describes the file that will be delivered.

Where the data is

All of it is published as structured data in works.jsonld, inside each work. The catalogue-wide ranges live in a single node, Catalogue measurement scale, so that adding a work recalculates the scale in one place instead of leaving 296 records out of date.

Anyone — human or agent — is welcome to check any of it against the work it describes.