Bokusho And The Brush That Has To Be Aimed

By Bokusho.

Digital calligraphy usually fails in a very specific way.

It follows the path.

That sounds reasonable until you have held a real brush. A real brush does not follow a path like a loyal cursor with decorative ink attached. The handle moves. The shaft tilts. The wrist leads or rests. The tuft compresses, bends, sticks, rolls, opens, closes, drinks, dries, and remembers what just happened. The mark on the paper is not the path. It is the consequence of a body.

Bokusho exists because the usual shortcuts are not good enough.

Fonts do not own calligraphy. Image generators do not own readable Japanese text. A textured stamp following a vector path does not own brushwork. It owns a costume.

The problem is more interesting than that, and meaner.

The Current Spine

Bokusho starts from ordered stroke geometry. KanjiVG gives us the source-order stroke paths for Japanese characters, which means we can begin with actual glyph intent instead of hallucinated pseudo-writing.¹

That matters, but it is not enough.

Stroke order gives the skeleton. It does not give the hand.

The live Bokusho renderer is being built around this pipeline:

stroke-order geometry
-> inferred calligrapher posture
-> handle path
-> roll, pitch, pressure, and height
-> deforming bristle/contact state
-> pigment, wetness, height, and paper response
-> rendered mark

That pipeline is the whole argument. The renderer is not allowed to pretend the stroke centerline is the visual center of every mark. It is not allowed to make dry brush by throwing random noise at the alpha channel and calling the wound ancient. It has to account for how a brush is held, aimed, pressed, turned, lifted, and reoriented.

There is a lot of machinery hiding inside that sentence. Good. There should be. Calligraphy is not difficult because ink is black.

What The Old Instruction Was Saying

Historical brush instruction does not read like a graphics paper. That is not a flaw in the tradition. That is a flaw in expecting old craft language to wear a lab coat before we listen to it.

Chinese brush-holding doctrine preserves the five-finger or five-word grip: thumb, index, middle, ring, and little finger form a counterbalanced control system around the shaft.² Modern summaries keep repeating the same practical rules: hold with solid fingers and an empty palm, keep the brush upright for ordinary center-tip work, vary wrist support by character scale, and adjust grip height for control versus freedom.³⁴⁵

None of that is decorative etiquette.

It is control theory with better sleeves.

The old brush-method vocabulary is full of operational claims:

center-tip: keep the point alive inside the stroke body
hidden tip and visible tip: control how the tuft enters and leaves the mark
lift and press: change depth continuously
turn and fold: compress, redirect, and reorient through a curve or corner
pause, check, and release: make the transition part of the stroke, not a cap
force at the tip: transmit intent through the body into the contact point

Those phrases are not asking for reverence. They are asking to become state.

What The Research Already Knows

There is serious virtual-brush literature. Bokusho is not pretending to have invented bristles in a heroic little shed.

Chu and Tai’s virtual Chinese brush work models a three-dimensional deforming brush, including flattening, bending, bristle spreading from lateral friction, wet-brush plasticity, and paper resistance.⁶ Robotic calligraphy work has measured relationships between handle coordinates, brush height, footprint shape, lag, and mark direction.⁷ More recent physically motivated brush work also treats pressure and footprint shape as calibration problems rather than vibes with funding.⁸

That work matters. It gives us the floor.

But it also reveals the crack.

Most computational approaches ask what footprint a handle motion produces. That is necessary. It is not the whole art.

The calligrapher is not merely suffering the brush dynamics. The calligrapher is commanding them.

The best human strokes often happen with a brush wet and tight enough that the tuft remains cohesive. That cohesion does not make the brush less expressive. It makes the brush more aimable. A small roll or pitch can steer the splay fan through a curve. A slight pressure change can open the tuft without losing the point. Wetness becomes a control affordance, not just a darker deposit.

Watercolor brush mechanics describe the physical side plainly: liquid between hairs pulls them together, and pressure spreads them while releasing liquid.⁹ The old masters describe the craft side: keep the tip alive, turn the brush, lift, press, hide, expose, redirect.

Put those together and the missing primitive becomes obvious enough to be annoying:

handle pose
-> roll / pitch / pressure / height
-> cohesive tuft fan direction
-> contact footprint

The brush is not only deforming.

The brush is being aimed.

What Bokusho Is Building

The current Bokusho implementation has started cutting away the dead stamp model.

Brush styles can now name posture, grip method, motion source, grip height, handle height, shaft tilt, center-tip bias, pressure, elasticity, bristle spread, ink load, water load, dry-brush tendency, speed, rhythm, taper, wrist lead, press/lift profile, and rotation.

Those are not final sacred parameters. They are the first handles on the machine.

The renderer now separates:

intended stroke geometry
inferred handle motion
lagging contact position
deforming footprint dimensions
stateful child tuft lanes
pigment and wetness deposition

The child tufts carry offset, width, load, wetness, phase, and dry memory across each stroke. They can separate, skip, deplete, recover, and recombine. That is still not full hair-level physics. It is not supposed to be. It is the next coherent layer: enough internal anatomy for the mark to stop behaving like one soft ellipse with a guilty conscience.

The next cut is the important one:

roll and pitch steering
-> torsion memory
-> cohesive tuft fan
-> side-specific ink reservoir
-> contact lanes with their own lag

That is where Bokusho starts treating the calligrapher’s tiny rotational authority as first-class render state.

Why This Might Be New Enough To Matter

The surprising part is not that nobody has modeled brushes. People have.

The surprising part is that digital calligraphy still so often forgets the thing every competent teacher eventually beats into your hand: the brush point must be controlled. The tuft is not just a passive material. It is an instrument whose shape, wetness, and direction are actively managed.

Traditional instruction preserved that knowledge as embodied practice. Computer graphics preserved a different slice as deformation and footprint simulation. Robotics preserved another slice as handle trajectory and calibration.

Bokusho is trying to put those slices back into one machine.

Not because the machine wants to cosplay as an old master. Spare us. The old masters had enough trouble already.

Because if we want digital calligraphy that does not look like crap, the render state has to carry the part of the art that lives between the fingers and the paper.

The brush is a wet, steerable body.

Aim it, or admit you are drawing with a stamp.

References

KanjiVG, stroke-order SVG data by Ulrich Apel and contributors: https://kanjivg.tagaini.net/ ↩
Chinese Notes, “Holding and Moving the Brush”: https://chinesenotes.com/calligraphy_writing.html ↩
Cleveland State University Pressbooks, “Introduction to Scripts & Skills”: https://pressbooks.ulib.csuohio.edu/intro-to-chinese-calligraphy/chapter/introduction-to-scripts-skills/ ↩
Chine-Culture, “Holding the brush”: https://new.chine-culture.com/en/chinese-calligraphy/holding-brush ↩
Shodo Kanji, arm and wrist carriage methods: https://shodo-kanji.com/a3-1-2brush_arm.html ↩
Nelson S.-H. Chu and Chiew-Lan Tai, “Real-Time Painting with an Expressive Virtual Chinese Brush”: https://cse.hkust.edu.hk/VCB/ ↩
“Calligraphy Brush Trajectory Control by a Robotic Arm,” Applied Sciences 2020: https://www.mdpi.com/2076-3417/10/23/8694 ↩
“Physically Motivated Model of a Painting Brush for Robotic Painting and Calligraphy,” Robotics 2024: https://www.mdpi.com/2218-6581/13/6/94 ↩
Handprint, brush mechanics and watercolor brush behavior: https://www.handprint.com/HP/WCL/tech18.html ↩