A Window, a Lamp · EyeLab
The first lab in Art: a studio lit by a north window, with plaster geometric solids and a celadon jar on the table. The sketch on the drawing board is computed live by the engine and drawn stroke by stroke. Journey from the inverted image of the camera obscura to Alberti's window; use Dürer's strings to trace a cube onto the window and watch vanishing points settle on the horizon line; lift a cylinder above eye level and watch the ellipse turn from flat to round; compare linear perspective with Guo Xi's "Three Distances" and handscrolls on a landscape sandbox; drag a lamp to watch the three values and five tones glide across a plaster sphere, and click on reflected light to trace where it comes from; squint your eyes until the white in shadow matches the black under the lamp; and finally, arrange your own still life, block in the values, and explore five different drawing styles for the same scene.
Chapters
Pinhole and Window
Against the studio wall stands a camera obscura: a small pinhole in front and a frosted glass screen behind. Light from the still life on the table travels in straight lines through the pinhole, projecting an inverted image onto the screen. Over two thousand years ago, the Chinese text "Mozi" recorded this: "The shadow turns upside down because rays cross at the pinhole" — the image is inverted because light rays intersect at the aperture.
Enlarge the hole and the image grows brighter, but each point blurs into a spot the size of the opening; shrink it too much, and diffraction spreads the light out. Between the two lies an optimal aperture that yields the sharpest image.
Now flip the screen in front of the pinhole, and place your eye where the pinhole was: the screen turns into a window, and the image right-side up — exactly as Leon Battista Alberti wrote in 1435 in "On Painting": a painting is an open window. The exact point where each light ray traveling from an object to your eye pierces the window is its position in the drawing.
Around 1420 in Florence, Filippo Brunelleschi performed a famous experiment: he drilled a small peephole through a wooden panel depicting the Florence Baptistery, looked through the back while holding a mirror in front, and found the painting lined up seamlessly with the real building. There was only one condition: the eye had to stay at that exact point.
A Window
In his "Treatise on Measurement" (1525), Albrecht Dürer illustrated a "perspective machine": a hook fixed to the wall serves as the eye, a string stretches from the hook to a point on a lute, and a mark is made where it passes through a frame; point by point, the lute emerges inside the frame.
In the 3D studio, click corners of the cube: a red string stretches from the eye, leaving a red dot on the window; the dot appears simultaneously on the drawing board. With enough dots, the shape of the cube emerges on its own.
Turn on Vanishing Points: a set of parallel lines converges to a single point in the drawing, which is the perspective image of that direction. Vanishing points for all horizontal lines fall on the same line — the horizon line, which lies right at your eye level. Sit down, stand up, or climb a ladder: the horizon line moves with you.
Face the cube head-on: only the edges receding into the distance converge; horizontal and vertical edges remain parallel to the picture plane and thus parallel to each other — one-point perspective. Turn it at an angle: two sets of horizontal edges each have their own vanishing point — two-point perspective. When the picture plane is upright, vertical lines remain strictly vertical; tilt your head down (tilting the picture plane), and verticals begin to converge as well — three-point perspective. All three perspectives follow the exact same rule.
Secret of the Circle
A circle in 3D space, when projected in perspective, becomes an ellipse — a true ellipse, with smooth, rounded curves at both ends and no sharp corners. Every ellipse on the picture plane is computed precisely by the engine (the plane containing the circle undergoes a projective transformation, turning the circle's equation into an ellipse).
How "round" an ellipse appears depends on how far it is from the horizon line: slowly raise the cylinder, and its top surface flattens; at eye level, it collapses into a straight line; raise it further, and its underside comes into view. Thus, for a cylinder on a table, the bottom ellipse is rounder than the top — because it sits further from the horizon line. Every rim, shoulder, and belly of a jar follows this same progression from flat to round.
Turn on Circle Center: the projected center of the circle does not land in the exact center of the ellipse, but offsets toward the farther side — the nearer half of the circle is closer to you and appears larger. This is why when drawing cross-axes through an ellipse, the front and back halves are never equal.
Three most common mistakes: drawing pointed ends on the ellipse ("lemon shape"); drawing the cylinder base as a flat straight line ("flat bottom"); and tilting the ellipse while the cylinder stands upright (the major axis must remain perpendicular to the cylinder's central axis). Use the "Common Mistakes" button to overlay them onto the correct ellipse for comparison.
Another Way of Seeing
The studio table now holds a landscape sandbox: a towering main peak behind, near mountains running in front, a hidden valley with a thatched cottage tucked behind the near ridges, a level line of distant mountains at the far horizon, and a river flowing in the foreground.
First look with one fixed eye (linear perspective): the main peak is compressed quite low, the valley is completely blocked by the near mountain, and the thatched cottage is invisible. This is a single instant captured in one window.
In "The Lofty Message of Forests and Streams", the Northern Song master Guo Xi wrote: "Mountains have three distances: looking up at the summit from below is high distance; peering into what lies behind from in front is deep distance; gazing across toward distant mountains from nearby ridges is level distance." The three distances represent three viewpoints: looking upward, peering down, and looking across. Stack them vertically into a hanging scroll, separated by drifting mist — and the same painting holds towering summits, views into hidden valleys, and vistas stretching to the horizon.
Now unroll the handscroll: drag the scroll, and your viewpoint moves along the riverbank, each section revealing what you see "upon arriving here". A handscroll is not a static window, but an unfolding journey.
This is not a failure to grasp perspective, but a different conceptual model: linear perspective records what one person sees from one spot at one single moment; the Three Distances and handscrolls record exploration — landscape that can be "traversed, viewed, wandered in, and lived in". Each serves its own purpose.
A Single Lamp
Drawing instructors frequently teach the "three values and five tones": light, halftone, and shadow; highlight, light halftone, core shadow, reflected light, and cast shadow. The zones on the drawing board are not pre-drawn, but computed from physics — drag the lamp (the glowing bulb in the 3D studio), and the five tones shift across the sphere.
They originate from just two sources: direct light from the lamp and reflections from the surroundings. Areas facing the light directly are brightest (light plane), growing darker as the angle steepens (halftone); turning to where light rays graze parallel to the surface, direct illumination vanishes — this is the core shadow (terminator). Further into shadow, the tablecloth, drape, and reflector bounce light back, brightening parts of the shadow — this is reflected light. Hence the core shadow is darkest: it receives neither direct illumination nor much reflected bounce.
Click anywhere on the drawing board, and the 3D studio reveals where its light originates. Remove the reflector and switch to a dark drape: reflected light almost completely vanishes. Now try making reflected light brighter than the light halftone — it is impossible, because reflected light is the product of a bounce and inevitably loses energy.
The cast shadow is darkest where the object meets the tabletop (where ambient light is nearly totally occluded), while its edges soften farther from the object; the larger the light source, the wider this soft blurred border (penumbra) becomes.
Squinting
A white cube sits in the shadow of a flag board, while a black-glazed jar stands under the lamp. Trust your intuition and guess first which is darker, then click twice on the board with the light meter to read their values (0 black, 10 white).
The human eye "corrects" what it sees: knowing the cube is white and the jar is black, it perceives their local color. But drawing paper holds only value — white in deep shadow can be darker than black under direct light. Drawing is about relationships, not an object's "original color".
The artist's trick is to squint: fine details dissolve, leaving only major value masses. Turn on "Squint" to simplify the scene into three or five value steps — this is the value plan made before laying down pencil strokes. In the eighteenth century, landscape painters carried tinted convex mirrors (Claude glasses) to compress tonal contrast and clarify major value relationships at a glance.
Finally, adjust the lamp until the value of the shadowed white cube matches the lit side of the black jar exactly: two objects — one white, one black — merge into the very same gray on paper.
Still Life Drawing
Select three or four plaster casts or objects from the shelf and drag them onto the table (drag objects inside the 3D studio). Next choose a viewpoint, position the lamp, and adjust framing with the viewfinder. The composition tips on the right are practical heuristics: avoid touching the paper edges with objects (edge tangents), avoid butting two silhouettes right against each other, and do not feel obliged to center the subject.
Once set, click "Start blocking in values": faint structural lines remain on the paper. Choose a pencil (one of the 3- or 5-step values) and block in broad tones on the drawing — click a region to fill it entirely, or click and drag to paint freely. Block in the darkest tones first, then midtones, leaving highlights white.
Click "Finish": the engine compares your tonal relationships against physical light simulation, evaluating primarily whether tonal relationships are correct (which region is darker than another), and secondarily absolute value precision. You can then let the engine render a complete drawing beside yours for comparison.
Likeness and Unlikeness
The same still life, same viewpoint, and same lamp rendered in five different artistic styles. Each is computed by the same engine, simply by omitting or transforming a different visual layer:
Full-tone drawing: captures form, lighting, and value comprehensively. Structural drawing: strips away most chiaroscuro to "draw through" hidden rear edges and central axes, focusing on spatial construction. Quick sketch: preserves only swift contour lines and a few strokes of shadow. Ink line drawing (baimiao): uses pure line, where modulated thickness and brushwork stand in for light and shade. Black-and-white woodcut: uses strictly black and white, carving marks following anatomical planes to create value steps.
These differ not in degrees of likeness, but in artistic choice: deciding what to preserve and what to discard. Accurate observation is the prerequisite of meaningful choice.
Qi Baishi said: "The subtle beauty of painting lies between likeness and unlikeness; too much likeness is vulgar, too little is deception."
In-class testing
Diagrams for each question are rendered live by the perspective and lighting engine; incorrect choices depict common drawing mistakes.
More labs
- VoltLab Circuit Lab
- Earth Motion Lab
- Cell · 3D structure
- Periodic Table · PeriodicLab
- Water Cycle 3D Sandbox · WaterCycle
- Neural Networks Reading Digits · NeuroLab
- From Switches to CPU · LogicLab
- Shenzhou Launch & Return · RocketLab
- We Were Here · ArtHistory
- Mendel's Pea Hybridization · PeaLab
- Humoral immunity · ImmuneLab
- The Cell's Energy Currency · ATPLab
- Solid Geometry · GeoSpace
- Contour 3D Lab
- Atmospheric Circulation Lab
- How Tides Form · TideLab
- Land and Sea Change
- Fluvial Landforms · RiverLab
- Geological Structures and Karst Sandbox · KarstLab
- Life of a Rock · RockLab
- Fly into the Eye · TyphoonLab
- Through the Earth · QuakeLab
- Microscope Bench · MicroLab
- The Structure of DNA · DNALab
- A Plant's Life · PlantLab
- The Stomach · StomachLab
- Cell Differentiation · StemLab
- Gene Expression · GeneLab
- Ecosystems · EcoLab
- Four Billion Years · DeepTime
- Preparing Gases in the Lab · GasLab
- Chemical Analysis Bench · AnalytLab
- Reaction Mechanisms · Molecular Animation
- Catalysts · CatalystLab
- Galvanic & Electrolytic Cells · ElectroLab
- Molecular Geometry · VSEPR
- Atomic Orbitals · Electron Cloud
- Chemical Reactions · The Micro World
- Conic Sections · ConicLab
- Function Graphs · FuncLab
- Compass & Straightedge · ConstructLab
- Pythagorean Theorem · PythagoLab
- Math Wonders · WonderLab
- Motion & Force · MotionLab
- Simple Machines · MachineLab
- Pressure & Buoyancy · FloatLab
- Thermodynamics · ThermoLab
- Heat & Changes of State · HeatLab
- Sound · SoundLab
- Optics Bench · OpticsLab
- Photoelectric Effect and Photons · PhotonLab
- Atomic Structure and Spectra · AtomLab
- Cloud Chamber & Nucleus · NucleusLab
- Oscillations & Waves · WaveLab
- Electric & Magnetic Fields · FieldLab
- Famous Battles Sand Table · WarTable
- Changing Territories · HistoryAtlas