How does 3D animation work?

Everything in a 3D film is hollow, like a balloon. Its skin is a mesh: a list of vertices (points in space), edges (straight lines joining two points) and faces (the flat tiles between the edges). Most film models are built from four-sided faces called quads.

A modeller often starts from a plain cube, cuts it into more quads, then pushes and pulls the points into a shape, a bit like sculpting clay. The result, the cage, is still blocky. So the computer runs subdivision: every quad splits into four and every point moves towards the average of its neighbours. Do it three times and 384 faces become 24,576, smooth as a pebble. Pixar first used this for a whole character in the short Geri's Game (1997).

The topology, meaning how the edges flow, matters as much as the count. Faces bend and stretch when the character moves, so modellers lay edge loops in rings around the eyes and mouth, the way the muscles run, so they can open, close and squint. Where the body bends, at elbows and knees, they add extra loops, so the skin can curve instead of folding like a paper straw.

Games must draw each frame in about 1/60 of a second, so their characters use tens of thousands of polygons. A film renderer can spend hours on a frame, so it chops the surface into micropolygons smaller than a pixel: millions for one close-up. The design itself usually comes from a turnaround drawing (see ConceptArtClear).

A mesh on its own is a statue. To make it move, a rigger builds a skeleton inside it: a chain of bones joined at pivots, just like your shoulder, elbow and wrist. Turn the upper-arm bone and everything below it follows.

Next comes skinning. Every vertex of the arm is told how much to follow each bone: its skin weights, which always add up to 100%. Near the elbow a point might follow the upper arm 50% and the forearm 50%, so the joint bends in a smooth curve. With hard 0-or-100% weights, the elbow folds like a paper straw. Artists paint these weights, and the heatmap shows them: red means 100%, blue means 0%.

Animators rarely turn bones one by one. With forward kinematics (FK) you rotate the shoulder, then the elbow, like posing a doll. With inverse kinematics (IK) you just move the hand, and the computer works out the joint angles, which is how a hand stays planted on a table. Faces use blend shapes: sculpted copies of the head (a smile, a raised brow) that a slider mixes in.

A film rig can have hundreds of these controls. Woody in Toy Story (1995) had 723, and 212 of them were in his face.

An animator doesn't draw every frame. They pose the rig at a few important moments, called keyframes, and the computer fills in the frames between (the in-betweens). Film runs at 24 frames a second, so this 2-second double hop is 48 frames, from just 5 keys.

How the computer fills the gaps is set by the interpolation. Stepped holds each pose until the next key, like a flip book of key drawings. Linear moves at constant speed, which looks robotic. Bezier follows smooth curves you shape with tangent handles, so a move can ease in and ease out, slowing down near a key.

The graph editor plots a value (here, hop height) against time. Its slope is the speed. A real hop is slow at the top, where the curve is flat, and fastest at the ground, where the curve should hit sharply like a bouncing ball.

Studios work in passes. Blocking sets the key poses in stepped mode to check the story. Splining switches to curves. Polish adds the details: squash and stretch, follow-through in the arms, tiny overlaps. These are the same principles 2D animators use (see Anim2DClear). A Pixar animator might finish only a few seconds of film a week.

A grey model needs colour and detail. You can't easily paint a curved 3D surface in a flat picture, so artists first unwrap it. They cut the mesh along seams and flatten the pieces into a square, like peeling an orange and pressing the peel flat. This flat map is the UV layout: every vertex gets a 2D address, u across and v up.

Now a painter works on the flat picture, the texture, and it wraps back onto the model through those addresses. A checker pattern is the classic test: if the squares stay square on the model, the unwrap is good. Film textures are huge, often 4K (4,096 × 4,096 pixels) or bigger, and one character may use hundreds of them.

Then shading decides how light behaves on the surface. Roughness spreads a highlight from a sharp sparkle (glossy) to a soft sheen (matte). Metalness makes reflections take the metal's colour. Subsurface scattering lets light soak into the material, bounce around and come out nearby, which is why skin, wax and marble glow softly at the edges. Rendering all this is the job of CGIClear.

An animator keys Golu's body, but nobody keys every fold of a cape. Things with thousands of moving parts, like cloth, hair, water, smoke, falling rubble and sparks, are simulated: the computer applies the laws of motion to each part, many times a second.

This cape is 192 points joined by invisible springs. Every step the computer moves each point by its speed and gravity, then pulls neighbours back to the right distance, over and over, and pushes points out of the body. The hair works the same way, with springs that try to keep each strand's shape. The blocks are rigid bodies with mass and spin. The sparks are particles, each born, falling and fading.

Keying this by hand would mean placing every point on every frame: this small scene alone would need hundreds of thousands of positions for ten seconds. So sims are computed, then an artist directs them with settings like wind, stiffness and gravity. Sulley in Monsters, Inc. (2001) had over 2.3 million simulated hairs, and Disney wrote a new snow simulator for Frozen (2013).

A 3D film is made like cars on an assembly line, shot by shot. In layout, artists place the camera, the sets and rough stand-ins, a 3D version of the storyboard. Animation keys the performance. FX adds simulations: dust, sparks, cloth, hair, water. Lighting places virtual lamps for mood. Rendering turns each frame into a finished picture. Compositing layers the rendered passes, adds glow and colour, and delivers the frame.

Rendering is huge. Toy Story (1995) has 114,240 frames and took about 800,000 machine-hours, roughly 7 hours a frame; Pixar could finish less than 30 seconds of film a day. Computers got faster, but films got richer, so frames still take hours. Monsters University (2013) averaged about 29 hours a frame on a render farm of some 2,000 computers. How renderers work is in CGIClear.

Toy Story was made by about 110 people. A studio feature today needs several hundred. India's first full-length 3D animated film, Roadside Romeo (2008), was animated in Bengaluru by about 150 artists over two and a half years. India's animation, VFX and post-production industry earned about ₹10,500 crore in 2025 (FICCI-EY), much of it working on films from around the world.