Round 05 · fluid slosh library

Thirty ways to move one measured pigment body

Hero 4A, its eight-fps clock, and the shared wall stay fixed. Palette noise is removed so the missing variable—isolated fluid behavior—can be judged directly.

Outcome: 10 promoted · 20 retained learnings
Active build time: 2h 06m 11s
Scope: 30 isolated loops · 270 direct paintings

Question

Which gesture has enough life without rebuilding the mud?

The first family prioritized drama and became too dense. The second met the ink budget but repeatedly collapsed into one crescent grammar. The third used the canonical in-repo Blue swatch as material truth and changed topology instead of asking for generic wet paint.

Family A · full wash. 35.3% average coverage; useful drama, knowingly too heavy for twelve-color stacking.
Family B · sparse perimeter. 12.9% average coverage; ink budget solved, motion grammar too repetitive.
Family C · source-swatch mechanics. 28.2% average coverage; canonical material and ten mechanically different kinds of flow.

Fixed control + source truth

Keep the winning score; isolate the missing motion

Accepted Round 04 control

Hero 4A · 3C cadence control

The three-second composite stays untouched while isolated mechanics change.

Canonical Little Plains Blue watercolor swatch — Material source of truth · canonical Blue, not a generic watercolor prompt.

Complete source library

Dense, sparse, and mechanically distinct

All thirty round-five neutral watercolor slosh atlases — All thirty attempts remain in order so the two failure families are legible beside the promoted family.

Family A · retained density study

Full wash · maximum gesture

35.3% average coverage

The first ten prioritize drama and silhouette. At 35.3% average coverage they are knowingly too ink-heavy for a twelve-color composite, but they preserve useful motion ideas.

Full wash · maximum gesture complete sprite atlas contact sheet — All ten full wash · maximum gesture atlases before motion review.

Full wash · maximum gesture states overlaid on the shared wall — Retained density study · combined geometry proof.

Study R5-01

Pendulum rebound

Retained density study

Playback: 8 fps
States: 9 direct

A heavy side-to-side bank with overshoot and recoil.

Study R5-02

Whirlpool scoop

Retained density study

Playback: 8 fps
States: 9 direct

A broad rotating scoop tightens and releases.

Study R5-03

Crown spill

Retained density study

Playback: 8 fps
States: 9 direct

A low basin rises into a crown and falls back.

Study R5-04

Double-wave chase

Retained density study

Playback: 8 fps
States: 9 direct

Two connected reservoirs chase one another.

Study R5-05

Impact rebound

Retained density study

Playback: 8 fps
States: 9 direct

A pooled body impacts the wall, curls, and rebounds.

Study R5-06

Barrel-roll breaker

Retained density study

Playback: 8 fps
States: 9 direct

A breaking wave rolls into a dense collapse.

Study R5-07

Asymmetric tumble

Retained density study

Playback: 8 fps
States: 9 direct

An uneven C-shaped mass tumbles around its weight.

Study R5-08

Undertow counterflow

Retained density study

Playback: 8 fps
States: 9 direct

A broad surface sweep meets an opposing undertow.

Study R5-09

Figure-eight exchange

Retained density study

Playback: 8 fps
States: 9 direct

Two reservoirs trade weight through an S-shaped waist.

Study R5-10

Centrifugal bloom

Retained density study

Playback: 8 fps
States: 9 direct

A compact deposit blooms outward and recoils.

Family B · retained similarity study

Sparse perimeter · minimum ink

12.9% average coverage

The second ten drop to 12.9% average coverage. They prove the layering budget, but the prompt collapsed too often into variations of one crescent grammar, so the whole family stays visible as evidence.

Sparse perimeter · minimum ink complete sprite atlas contact sheet — All ten sparse perimeter · minimum ink atlases before motion review.

Sparse perimeter · minimum ink states overlaid on the shared wall — Retained similarity study · combined geometry proof.

Study R5-11

Perimeter ribbon chase

Retained similarity study

Playback: 8 fps
States: 9 direct

A narrow weighted ribbon runs around the wall.

Study R5-12

Slingshot hook

Retained similarity study

Playback: 8 fps
States: 9 direct

A pooled head stretches a tail and snaps around.

Study R5-13

Shallow tide basin

Retained similarity study

Playback: 8 fps
States: 9 direct

A low pool banks left, crosses, and banks right.

Study R5-14

Whip curl

Retained similarity study

Playback: 8 fps
States: 9 direct

A slender hook unfurls, curls, and recoils.

Study R5-15

Counterbalanced crescent

Retained similarity study

Playback: 8 fps
States: 9 direct

A narrow band trades pooled weight between its ends.

Study R5-16

Orbiting tongue

Retained similarity study

Playback: 8 fps
States: 9 direct

A tapered tongue chases a compact pool around the rim.

Study R5-17

Squeeze-release ribbon

Retained similarity study

Playback: 8 fps
States: 9 direct

A thin arc compresses against opposing walls.

Study R5-18

Serpentine counterflow

Retained similarity study

Playback: 8 fps
States: 9 direct

A narrow S-wave bends in opposing directions.

Study R5-19

Rolling eyelid

Retained similarity study

Playback: 8 fps
States: 9 direct

A shallow arc rolls from base to crown and back.

Study R5-20

Three-beat eddy

Retained similarity study

Playback: 8 fps
States: 9 direct

A hooked pool curls, settles, and curls again.

Family C · canonical material

Source-swatch mechanics · different kinds of flow

28.2% average coverage

The final ten use the in-repo Blue swatch as the exact material source, then test folding, transfer, jets, barrels, bridges, fountains, pinches, coils, kneading, and collision. Only the hue is neutralized after generation.