Suminagashi Generative Art With Advection

Suminagashi generative art can study floating rings, breath-like perturbation, and transferred contour through a computational field. It should not collapse a centuries-old Japanese marbling practice into a shader aesthetic or imply that simulated particles reproduce ink, surfactant, water, paper, gesture, and craft.

This field note names the digital translation precisely: deposit concentric rings, advect their points through a bounded velocity field, preserve a mass-like ledger, and compose the final still as an edition with cultural and algorithmic provenance.

Suminagashi generative art needs cultural specificity

Suminagashi is commonly translated as “floating ink.” Its history, materials, repeated ink and dispersant touches, water movement, and paper transfer matter. A browser study can learn from concentric rings and delicate disturbance, but it should describe what it models and what it leaves out. Credit the practice in the title, caption, and project notes rather than using it only as a style tag.

The Metropolitan Museum's “Playing with Paper” essay gives collection-based context for paper decoration and marbling. Read objects, dates, places, materials, and making descriptions before building a palette. Suminagashi generative art should not claim a specific historical motif from a generic turbulent image.

Define the digital translation as rings represented by points or pigments on a plane, moved by a velocity field and rendered as a still. Call it a floating-ink simulation only with the qualifier that the field is illustrative, not a validated fluid model. The worked artifact conserves point count and bounds radial growth; it does not establish physical mass conservation, surface tension, or paper absorption.

Deposit rings as events, not anonymous pixels

Represent a ring by deposition ID, center, radius, sample order, ink role, creation tick, and parent touch. Sample around angle with enough points for output curvature, then preserve circular order through advection. Multiple depositions can alternate dark ink and a separating role, but label this as a compositional analogue unless actual material behavior is modeled and cited.

The Met's marbled paper essay offers another museum perspective on paper marbling traditions and objects. Use museum records as provenance, not as a parameter catalog stripped of context. Suminagashi generative art should retain source notes beside code and distinguish historical observation, contemporary interpretation, and algorithmic invention.

Choose deposition rhythm deliberately. “Pulse” alternates tight rings at one center. “Conversation” uses two centers whose rings meet.

“Drift” places a slow diagonal series that the field carries together. These are visual studies, not reconstructions of named historical works. Keep a pre-advection still for each so the later marbling flow field can be judged by how it transforms known structure rather than by complexity alone.

Runnable artifact: The ring-and-advection fixture checks point-count preservation, finite motion, a zero-step identity, bounded radius, empty input, repeatability, and a hard large-step control.

Save this worked fixture as suminagashi-advection.test.mjs and run node suminagashi-advection.test.mjs. Expected final line: PASS: 10 ink assertions.

import assert from "node:assert/strict";
const ring=(count,radius)=>Array.from({length:count},(_,i)=>[radius*Math.cos(i*2*Math.PI/count),radius*Math.sin(i*2*Math.PI/count)]);
const advect=(points,dt)=>points.map(([x,y])=>[x+dt*(-y*.08),y+dt*(x*.08)]);
const mass=points=>points.length;const maxR=points=>Math.max(...points.map(([x,y])=>Math.hypot(x,y)));
let n=0;const check=fn=>{fn();n++};const seed=ring(32,10);const moved=advect(seed,.25);
check(()=>assert.equal(mass(seed),32));check(()=>assert.equal(mass(moved),32));
check(()=>assert.ok(maxR(moved)<10.01));check(()=>assert.deepEqual(advect([],1),[]));
check(()=>assert.deepEqual(advect(seed,0),seed));check(()=>assert.notDeepEqual(moved,seed));
check(()=>assert.ok(moved.flat().every(Number.isFinite)));check(()=>assert.equal(ring(1,10)[0][0],10));
check(()=>assert.deepEqual(advect(seed,.25),moved));check(()=>assert.ok(maxR(advect(seed,100))<100));
assert.equal(n,10);console.log("PASS: 10 ink assertions");

Advect with a field you can draw

Define velocity as a sum of bounded components: broad translation, one or two vortices, gentle divergence, and perhaps a low-amplitude seeded curl field. Draw arrows or streamlines for a sparse grid beside the output. A field becomes art-directable when a dense fold can be traced to a vortex position or pulse, not when noise is added until the rings look watery.

Use a fixed timestep and a named integrator. Compare Euler against a midpoint or Runge–Kutta control on a small ring, especially near strong curvature in the field. Cap velocity and displacement per step. Surfactant advection is a physical phrase with specific implications; if the implementation advects line samples through an invented vector field, state that it is an artistic analogue rather than a surfactant solver.

Flow-field plotter art provides a useful visual and numerical discipline: seed, integration, boundary, stopping, and path budgets determine the drawing. Suminagashi generative art adds nested deposition order and the expectation that nearby ring points remain coherent. Track local segment stretch and orientation flips so a dramatic field does not turn contours into accidental self-crossing noise.

The still-image control is reviewed before texture. If nested contour and directional breath are absent there, paper grain cannot rescue the composition.

Track mass-like quantities without claiming physics

For a polyline representation, point count, ring count, segment length, enclosed signed area, and out-of-bounds points are useful computational summaries. They are not ink mass. If raster pigment is modeled, track total scalar pigment before and after advection, diffusion, deposition, clipping, and transfer. Name boundary losses and numerical tolerance.

The computational marbling preprint offers a current research reference for digital simulation. A local implementation should state which equations or ideas it actually adopts. Suminagashi generative art can remain intentionally simpler: a geometric advection study whose evidence is repeatability, bounded deformation, and a clear separation between source rings and rendering.

Plot metrics over ticks for one small fixture. A sudden path-length explosion may signal an unstable timestep. A shrinking enclosed area may be intended compression or clipping.

A ring-order reversal may indicate a fold worth preserving or a numerical defect, depending on the visual contract. The receipt should distinguish measured behavior from interpretation. That restraint makes the later paper texture and ink tone more believable, because they are not asked to disguise an unknown simulation.

Render paper as a receiving surface

Keep geometry and surface separate. Draw a flat one-color control first. Then add paper tone, subtle fiber-scale variation, edge softness, and ink density with bounded effects.

Do not feed the textured raster back into advection. Canvas blend modes can model layered visual interactions, but each blend should have a normal-composite comparison and a reason connected to absorption, overlap, or tone.

Digital paper marbling can be vector, raster, plotted, or printed. Vector contours retain traceability and scale; raster fields can express diffusion and granular tone; a plotter introduces real ink and paper but not floating transfer. Describe the medium honestly. A giclée print of a simulation is not a marbled sheet, and that distinction does not reduce its value as an image.

Crop after the complete field. Preserve quiet water as negative space rather than filling every corner. A title can align with the field's broad flow while remaining outside the most delicate feathers.

Test color in grayscale and on the chosen paper profile. Suminagashi generative art benefits from near-black, one restrained accent, and paper warmth, but the edition should be able to explain palette as interpretation rather than inherited authenticity.

Bound motion and preserve a meaningful still

An interactive version can show deposition and advection step by step, but motion should not be the only way to understand the piece. Respect reduced motion with the final still and optional manual stepping. Expose pause, reset, seed, field name, and tick. Do not autoplay rapid swirling that contradicts the delicacy of the visual source or creates unnecessary vestibular load.

Set budgets for rings, points, ticks, velocity evaluations, canvas pixels, export bytes, and main-thread time. Move heavy raster work off the main thread where appropriate, but preserve the same deterministic state. Reaction-diffusion WebGPU art uses a different field evolution; compare its scalar chemistry with the contour transport here instead of merging both into generic organic motion.

Accessibility text should describe deposition centers, dominant direction, nesting, compression, feathering, palette, and negative space. A small semantic table can list field components and metrics. If canvas fails, show the verified static SVG or image. The reader should retain the thesis—rings stretched by a bounded field—even without animation, texture, or color.

Publish context, algorithm, and surface receipts

Archive cultural sources, wording decisions, deposition events, seed, ring sampling, field equations, component parameters, timestep, integrator, boundary, tick count, deformation metrics, renderer, texture layers, crop, palette, output, accessibility description, and code revision. Include source rings, a field diagram, three matched variants, a flat control, and a physical proof.

Seeded randomness in generative art helps hold field noise and deposition proposals stable. Reject an edition that changes the seed between variants, loses ring identity, uses unstable steps, clips without disclosure, relies on motion for meaning, or treats Japanese craft as an exotic texture label. Suminagashi generative art deserves both technical and cultural review.

The final image should feel suspended rather than merely distorted. Nested contours remember where ink touched, the field carries them into a broad breath, and paper tone gives the movement somewhere to settle. Its receipt states the limits plainly: this is a computational study informed by a living material tradition. That honesty leaves room for both engineering precision and genuine visual reverence.