Facial Animation for Games: Expressions, Lip Sync and Visemes

Facial animation is where a game character stops being a moving model and starts being someone. Players forgive a stiff walk long before they forgive dead eyes or a mouth that flaps out of time with the voice. This guide explains how facial animation works in games, from the rig under the skin to expressions, lip sync and visemes, and what to practise first if you want to animate faces yourself.

What facial animation is

Facial animation is the movement of a character's face: brows, eyelids, eyes, cheeks, nose, lips, jaw and tongue. It has two jobs. The first is acting, where the face shows what the character thinks and feels. The second is speech, where the mouth matches the dialogue so the voice looks like it comes from the character.

In games, faces show up in three places, and each asks for a different level of quality:

• Cinematics and cutscenes. The camera is close and the player is watching, not playing. This is where hand-keyed or captured facial performance gets the most attention.
• In-game dialogue. Conversations with other characters during play, often hundreds or thousands of lines. Quality matters, but so does volume.
• Gameplay reactions. Pain, effort, surprise, blinking and looking around. Short, reusable animations that keep the character alive between story moments.

Facial work sits between two specialisms. Animators perform the face; technical artists build the facial rig that makes the performance possible. On the Certificate in 3D for Games, those are two of the four specialisations: Animation, taught in Maya, and Tech Art, taught in Maya and Python. If faces interest you, it is worth knowing early which side of that line pulls you more.

How a facial rig works

You cannot animate a face without a rig. A facial rig is the set of controls an animator moves, and the system underneath that deforms the mesh. There are two main approaches, and most game characters use a mix of both.

Joint-based facial rigs

Here the face is driven by small joints (bones) placed around the brows, eyelids, cheeks, lips and jaw. Each joint pulls on the nearby vertices through skin weights, the same system used for the body. Joint rigs are cheap to run in real time and export cleanly to game engines. Their weakness is fine detail: creases, lip roll and skin sliding over the skull are hard to get from joints alone.

Blend shape rigs

A blend shape is a sculpted copy of the face in a specific pose, such as a smile, a blink or a closed "M" mouth. The rig blends between the neutral face and these shapes. Maya calls them blend shapes, Blender calls them shape keys, and Unreal Engine calls them morph targets. They are the same idea. Blend shapes give you exact control over how the face looks in each pose, which is why they are popular for hero characters. They cost more memory and need careful sculpting, and combining shapes can produce ugly results unless the rig adds corrective shapes.

Building on FACS

Most high-end facial rigs for realistic characters are organised around FACS, the Facial Action Coding System. FACS breaks every facial expression into small muscle movements called action units, such as the inner brow raiser, the cheek raiser or the lip corner puller. A rig built on FACS gives the animator one control per movement instead of a fixed set of whole expressions, so you can build any expression from parts. It also makes rigs compatible with facial capture data, which is usually solved into the same kind of units.

For skin weights, controls and export to an engine, see our guide to rigging a character for games.

Animating expressions

Good facial animation is acting. Before you touch a control, know what the character is thinking in the shot, and what changes in their thinking. The face should show those changes, not decorate every word.

Start with the eyes

Players look at eyes first, so animate them first. Three things matter most:

• Eye direction. Where the character looks tells the player what they are thinking about. A glance away before a lie reads instantly.
• Eye darts. Real eyes make tiny, fast jumps rather than smooth drifts. A few small darts make a held pose feel alive.
• Blinks. Use them with purpose. A blink often marks a change of thought or happens as the head turns. A blink every two seconds on a timer looks robotic.

Remember that the lids follow the eyes. When the eye looks down, the upper lid comes down with it. Rigs often automate this, but check it, because lids that ignore the eyeball make a face look like a mask.

Brows, cheeks and the mouth corners

After the eyes, the brows carry most of the emotion. Then the mouth corners and cheeks. A real smile lifts the cheeks and narrows the eyes; a smile that only moves the mouth reads as polite or forced. Build expressions from the top of the face down and check each one as a still frame. If the pose does not read without motion, motion will not save it.

Asymmetry and restraint

Real faces are rarely symmetrical. Offset the brows slightly, let one mouth corner lead the other, stagger the timing of the two sides by a frame or two. Then hold back. Beginners tend to animate a new expression on every word. Fewer, clearer changes look more confident and more human.

Lip sync: making the mouth match the voice

Lip sync is the part of facial work beginners overcomplicate most. The key idea: animate the sounds you hear, not the letters in the script. "Knight" starts with an N sound, not a K. Listen to the line many times and mark the important sounds on the timeline first.

Phonemes, visemes and the mouth chart

A phoneme is a unit of sound. A viseme is the mouth shape that goes with it. Many phonemes share one viseme, because you cannot see the difference between them on the lips. P, B and M all look the same: lips pressed together. That is why lip sync works with a small set of mouth shapes rather than one shape per sound.

A typical animation mouth chart groups the shapes like this:

• M, B, P: lips closed and pressed.
• F, V: lower lip tucked under the top teeth.
• O, OO, W: lips pushed forward into a round shape.
• EE and other wide sounds: mouth corners pulled out, teeth close together.
• AH, open vowels: jaw dropped open.
• L, TH: the tongue is visible, touching the top teeth or poking between them.
• Everything else: a relaxed, slightly open rest shape that covers many consonants.

Lip sync tools use their own viseme sets, often somewhere between six and twenty shapes, and the names vary. Many modern game rigs skip visemes and drive the FACS controls directly, but the mouth still has to hit the same key shapes.

A lip sync workflow that works

1. Jaw first. Animate only the jaw opening and closing on the syllables. This gives the rhythm of the line. Do not open the jaw fully on every syllable; follow the stressed ones.
2. Width second. Add the mouth corners moving wide and narrow, for sounds like EE and OO.
3. Contacts third. Make sure every M, B and P closes the lips completely, and every F and V touches lip to teeth. These are the shapes viewers notice if you miss them.
4. Tongue last. Add the tongue for L, TH, T and D where it is visible. On distant game cameras you can often skip it.
5. Shift it early. Mouth shapes usually look in sync when they arrive a frame or two before the sound. Try nudging your keys earlier and compare.

The most common mistake is the "chattering" mouth: a full open and close on every syllable, so the jaw flaps like a puppet. Real speech blends sounds together. Drop the small sounds, keep the accents, and let the mouth stay partly open through fast phrases.

How games produce facial animation at scale

A story-heavy game can have more dialogue than a team can keyframe by hand on schedule. So studios split the work by importance.

• Hero cinematics are usually captured with facial performance capture or keyed by hand, then polished by animators. Capture gives a strong base, but it still needs cleanup: fixing lip contacts, adding eye darts, pushing expressions that came through too weak.
• Standard dialogue is often generated from the audio file by lip sync software, then given expressions and eye movement on top, either by hand or from a library of reusable facial animations.
• Barks and reactions, the short lines and grunts during gameplay, rely mostly on automatic lip sync and stock expressions.

Generated lip sync saves time, but it does not act. It matches the mouth to sounds; it does not know the character is lying, scared or joking. That is why animators who understand facial acting stay valuable even when the mouth is automated.

Real-time limits shape the work too. A face that the camera only sees from a distance does not need every blend shape, and characters often switch to simpler facial rigs at lower levels of detail. In Unreal Engine, facial animation usually arrives as morph target curves, joint animation, or both, so the rig and the animation must agree on the same names and setup. This is where animators and technical artists work closely together.

What to practise first

If you are new to facial animation, keep the scope small:

1. A blink and a look. A character looks at something, blinks, and looks back. It teaches eye timing and lid follow-through.
2. An expression change. Neutral to surprised, or happy to worried, with no dialogue. Focus on the thought change and the order the face moves.
3. A short line of dialogue. Pick a clip of two to five seconds with clear emotion. Do the jaw, width, contacts, tongue pass described above.
4. A dialogue shot with acting. Combine lip sync, eyes, brows and a little head movement. This is the kind of shot that belongs in a reel.

Use a free, pre-made character rig with a decent face, so you practise animating rather than rigging. Shoot reference of yourself saying the line. You will see how little the jaw actually moves.

What studios look for

For animators, studios want to see clear acting: a face that thinks, eyes that lead, and lip sync with clean contacts and no chattering. One short dialogue shot that holds up frame by frame beats a long reel of busy mouths. Show the body as well; a great face on a frozen body will not convince anyone.

For technical artists, the evidence is a working facial rig: clean deformation, sensible controls an animator can use quickly, and proof that it runs in a game engine. A short video of the rig moving through extreme expressions and a few lines of lip sync shows more than any description.

Either way, the foundations come first. Facial animation builds on solid body animation and on timing and the other principles covered in our guide to the 12 principles of animation. Faces are rarely the place to start.

Where to go from here

If you are new to 3D, start with the basics. The free Blender 3D Bootcamp is six live online sessions over three weeks, with no experience needed, and it ends with a session in Unreal Engine. It gives you the footing that facial work depends on.

If you are ready to commit to games, the Certificate in 3D for Games is a 60-week online programme where you try each specialisation before choosing between 3D Art, Tech Art, Animation and Procedural. Animation and Tech Art are the two routes into the work this guide describes, and you put them to work in team projects each term, ending in a vertical slice.