What Is Cel Shading? How 3D Learns to Draw
Updated: 1 day ago
Published September 13, 2026
Cel shading is a non-photorealistic rendering technique that makes three-dimensional models read like drawn animation. Its central trick is not simply adding a black outline. It replaces smooth, physically plausible changes in light with deliberate graphic decisions: a bright shape, a shadow shape, perhaps a highlight, and carefully controlled edges. The result preserves 3D movement and camera freedom while borrowing the visual grammar of comics, animation cels, posters, and ink. Cel shading works when every layer agrees about what information matters—and fails when a normal 3D scene is merely flattened.
That distinction matters because cel shading is often described as a filter. It is better understood as a small visual system. The shader, model, textures, line work, lighting, animation, and camera all negotiate the same question: which parts of this form should the eye read first?
Cel shading versus flat shading
Flat shading and cel shading can look related, but they solve different problems. Flat shading usually assigns one surface normal to each polygon, so every polygon displays a uniform tone and the mesh’s facets become visible. A low-poly crystal rendered this way looks angular because its faces do not blend.
Cel shading can be applied to a smoothly modeled surface. The renderer still calculates how a surface faces the light, but instead of displaying a continuous gradient it maps that calculation into a small number of tone bands. A cheek may have one broad light region and one sharply bounded shadow region even though the underlying geometry is curved.
So flat shading exposes polygon structure; cel shading edits the vocabulary of light. Either can appear without a black contour. Either can be combined with the other. Confusing them makes it harder to see what the artist actually designed.
The core mechanism: turn a gradient into decisions
In conventional diffuse shading, a surface usually grows brighter as its normal points more directly toward a light. The change is continuous. Cel shading begins with the same geometric relationship, then refuses most of the intermediate values.
Imagine the calculated brightness as a number from zero to one. A simple toon shader might send values below 0.45 to a shadow color and everything above that threshold to a light color. A three-band version could add a middle tone. Another approach samples a small ramp texture, allowing the artist to paint where tones change and what colors they become.
This is why the influential 2000 paper Stylized Rendering Techniques for Scalable Real-Time 3D Animation matters: it treated stylization as something real-time 3D rendering could generate coherently and at multiple levels of detail, rather than as a painted effect pasted onto a frozen image.
The threshold is simple mathematics. The art begins with everything arranged around it.
Six layers that create the illustrated illusion
1. Tone bands
The number, color, and placement of light bands establish the basic graphic language. Two hard bands can feel like cut paper. A soft transition between two broad regions can resemble digital anime compositing. Colored shadows can preserve mood without returning to realism.
Earlier non-photorealistic research also challenged the assumption that realistic luminance is always the clearest way to describe form. The 1998 Gooch lighting paper used warm-to-cool color shifts and reserved extreme lights and darks for highlights and edges. It was aimed at technical illustration rather than anime, but the principle travels: rendering can clarify selected structure by refusing photographic light.
2. Contours and internal lines
Outlines are optional, but they are powerful. A silhouette line separates a character from the background. Crease lines describe folds or mechanical joints. Material boundaries can mimic inked divisions. The line may be created from geometry, screen-space edge detection, depth and normal differences, or a dedicated line renderer.
Blender’s official Freestyle documentation describes an edge-and-line-based non-photorealistic renderer that uses mesh and depth information to select features. That is a useful reminder: “the outline” is not one operation. Artists decide which edges deserve a mark, how thick it should be, and whether the line changes with distance, lighting, or importance.
3. Authored normals
A physically ordinary normal follows the model’s local surface. A stylized production may edit or replace normals so light wraps across hair, faces, or clothing in cleaner shapes. Geometry can be lumpy while the shadow behaves like a confident brushstroke. This is not an error in the model. It is the digital equivalent of an illustrator ignoring a minor bump to preserve the larger form.
4. Texture and material control
Skin, cloth, metal, hair, and eyes should not all obey the same rule. A face may suppress specular highlights; an eye may place a designed highlight that barely responds to the environment; metal may need a second ramp; hair may use painted masks to stop shadows from breaking into noise. Good cel shading is selective, not universally flat.
5. Animation timing and deformation
Perfectly smooth interpolation can betray an otherwise drawn-looking frame. Traditional animation selects poses and controls the spacing between them; real-time 3D normally reveals every intermediate position. A stylized game may hold poses, step parts of an animation, distort the mesh for a particular camera angle, or change a facial feature between designed shapes.
Arc System Works’ own account of its Guilty Gear Xrd GDC presentation describes a deliberate effort to rebuild a classic 2D fighting game inside a modern 3D framework while keeping its old-school 2D character. That result was not produced by a shader alone; it came from coordinated art, programming, modeling, animation, and camera decisions.
6. Camera and composition
A style that looks convincing from one angle can collapse when the camera moves. Silhouettes overlap. A nose becomes too dimensional. An outline exposes the wrong crease. This is why fixed or constrained cameras make some highly authored styles easier, while free-camera games need rules that remain readable from many views.
Nintendo’s Wind Waker developer interview is valuable because it discusses visual choices alongside animation, proportions, hardware, and the game’s world. The enduring look is not reducible to one rendering checkbox. It is a whole design language built around expressive motion and simplified form.
A brief history: from NPR research to visual identity
Computer graphics did not discover stylization after finishing realism. Researchers in the 1990s were already building non-photorealistic rendering systems for illustration, silhouettes, technical diagrams, painterly marks, and animation. The category’s rebellious premise was practical: a photographically plausible image is not automatically the most informative or expressive image.
Around the turn of the millennium, real-time hardware made bold stylized rendering increasingly visible in games. Jet Set Radio became an early public emblem of polygonal worlds that looked drawn. The Wind Waker demonstrated that a major adventure could make simplified color and exaggerated animation central to its identity. Later projects pushed the technique beyond broad flat bands into sophisticated systems that manipulate normals, pose timing, facial geometry, effects, and compositing.
The timeline is not a march from primitive cel shading to advanced realism. It is an expanding toolbox. A modern production can choose crude two-tone shadows because crudeness is the point, or build an elaborate shader whose complexity disappears into apparent simplicity.
Why cel shading ages differently
Photorealistic graphics invite comparison with the newest simulation. Viewers notice yesterday’s hair, skin, reflections, and facial capture because the work promised a version of reality. Cel shading makes a different contract. Its success depends more on internal consistency, silhouette, color, timing, and composition—qualities that do not become obsolete merely because a newer graphics card can trace more rays.
That does not make stylization timeless by magic. Fashion, interface conventions, resolution, animation budgets, and cultural associations still date a work. The difference is that visible limitation can become part of the design instead of evidence that the simulation failed.
The same principle appears in Why Cyberpunk Cities Look Better in the Rain: visual elements work when they perform clear narrative jobs, not when they are piled on as genre decoration. A toon outline without a hierarchy of forms is just another effect competing for attention.
A field guide for reading cel-shaded art
When a cel-shaded game or animation looks unusually convincing, inspect it with six questions:
Where do tones change? Are shadow boundaries describing form, emotion, composition, or merely the light’s raw calculation?
Which edges are drawn? Does line weight separate silhouettes, materials, details, and depth—or outline everything equally?
What refuses physics? Look for painted highlights, edited normals, disappearing features, and shadows that protect the face.
How does motion behave? Are poses held, stepped, stretched, or reshaped to preserve a drawn rhythm?
How much freedom does the camera have? Is the image authored for selected views or robust from every angle?
Does the environment speak the same language? A stylized character inside conventionally realistic lighting often looks composited rather than present.
For another way interfaces and images become part of a game world rather than decoration around it, see What Is Diegetic UI? The shared lesson is integration: an effect becomes a language only when the rest of the design answers it.
The point is not to imitate paper
Cel shading is sometimes judged by how successfully it fools the eye into seeing a drawing. That can be an impressive goal, but it is not the only one. The stranger possibility is a hybrid image that could not exist as traditional animation or conventional 3D: drawn but navigable, graphic but spatial, mechanically lit yet visibly authored.
The technique succeeds when it stops apologizing for not being real. It turns light into shapes, edges into decisions, and technical limitations into style. The machine does not become an illustrator by hiding its mathematics. It becomes one by learning which mathematics to disobey.
Which cel-shaded game or animated 3D work best preserves the illusion of drawing for you—and is the decisive ingredient its lighting, outlines, animation timing, camera work, or something else?
If you enjoy graphics that turn digital artifacts into a visual language, explore the Glitchwear collection and join the Claw & Riot Salon to compare examples and arguments.
Comments