SFM Facial Animation: Flexes, Face Posing and Expressions

How Faces Work in Source Filmmaker

Body animation in SFM moves bones. Most facial animation uses flexes instead, known as shape keys or morph targets in other software. Each flex is a sculpted version of the face mesh, such as a raised inner brow, and a slider blends the neutral face toward it. Fan ports such as Fortnite SFM models vary a lot in how many face flexes they include, so check this before you plan close-ups. Flexes are built into a model when it is compiled, and our SFM compile guide explains that process.

Flexes are authored when the model is made. Source Film Making (SFM) cannot add new face shapes at animation time, so your range of expressions depends on what the modeler built.

Flexes and Flex Controllers

The slider you see is a flex controller, not always a single flex. One controller can drive several flexes, and corrective shapes can fire in the background to fix ugly overlaps when two shapes combine. You only ever touch the controller. SFM keeps each controller between 0 and 1, so you cannot push a shape past the strength the modeler gave it. Garry’s Mod faceposing can overdrive flexes. SFM cannot.

Single, Bipolar, Stereo and Multi Sliders

Face sliders do not all behave the same way. Knowing the type tells you where neutral sits before you drag anything.

Slider type	How it behaves	Example
Single direction	0 is neutral, 1 is the full shape	An eyelid close control
Bipolar (V or H sliders)	One shape at each end, neutral near the middle	LipsV, which closes the lips one way and opens them the other
Stereo	Can hold different values on the left and right side	Smile on HWM models
Multi	A selector picks between shapes and a second value sets intensity	multi_smile
Combination (preset-style)	One slider drives a whole expression	happyBig or mad on standard TF2 models

Combination sliders are quick but blunt. Individual sliders take longer and give you real control.

Finding the Face Controls in the Animation Set Editor

Expand the character’s animation set in the Animation Set Editor. You see a tree of control groups, and selecting a group selects every control inside it. The selected controls appear as sliders in the Controls panel. Above them, the Presets panel holds preset sliders that act on whatever you have selected.

If you select an eye or jaw item and see position and rotation values instead of a slider, it is a bone, not a flex. Bones are posed with the viewport manipulators.

Face Groups Differ by Model

Group names are set by the model, so read what your character offers rather than following a fixed list. The standard TF2 player models show a short face group with combination sliders such as happyBig, mad and painSmall. The HWM versions in the player/hwm/ folder split the face into many individual controls, including Lower Face and Tongue groups for the mouth, brow sliders such as BrowInV, and an Eyes group. Our guide to importing SFM models from the Steam Workshop covers where those models land and how to load them.

Model type	Face controls	Best use
TF2 HWM class models	Many individual flex sliders plus Emotion and phoneme presets	Dialog scenes and close-ups
Standard TF2 player models	A few combination sliders	Wide shots and quick reactions
HL2-style FACS models	Individual flexes following the Half-Life 2 layout	Detailed posing; phoneme presets can be generated
Custom or ported models	Whatever flexes the porter built	Check each slider before planning a scene
Bone-based faces	Face bones, no flex sliders	Pose with manipulators or bone presets

The Presets Panel and the Controls Panel

The two panels behave differently, and mixing them up causes confusion. A control slider stays where you drag it. A preset slider blends the selected controls toward a preset, then springs back. The Procedural tab is always there and holds tools such as Default, Smooth and Jitter. Character models add their own tabs, typically Emotion, Phoneme and Viseme. Valve explains the split on its Presets panel and Controls panel page.

For exact values, double-click a control slider and type a number.

Building an Expression From Brows, Eyes and Mouth

Before touching a slider, move the work camera close and check the pose from the angle your shot camera will use. Then work from the top of the face down.

Brows First

Brows carry more emotion than most beginners expect. Inner brows raised together read as worry or sadness. Brows pulled down and in read as anger or focus. One brow up reads as doubt.

Eyelids and Squint

Eyelids control intensity. Wide lids add fear or surprise. Lids lowered slightly add calm, boredom or contempt. Tightened lids with raised cheeks turn a polite smile into a real one. Look for lid close and squint controls on your model; their names and groups vary from character to character.

Mouth and Cheeks

The mouth confirms what the brows suggest. Lip corners up or down set the basic mood. The jaw adds energy: a slightly open jaw looks relaxed, a clenched jaw looks tense. For a closed-mouth smile, raise the smile control and leave the lip opening sliders near neutral so the teeth stay hidden. A smile without cheek movement looks pasted on.

Asymmetry With the L-R Slider

Perfectly symmetrical faces look like masks. The L-R balance slider in the center of the Procedural tab lets you apply a stereo control to one side more than the other. Drag it toward one side, then adjust the smile and you get a smirk. It only works on stereo controls, so some sliders ignore it, and on some models few controls are stereo at all.

Expression	Brows	Eyes and lids	Mouth and jaw
Happy	Slightly raised	Cheeks lifted near the nose	Corners up, jaw slightly open
Sad	Inner brows raised	Lids slightly lowered	Corners down, chin raised
Angry	Lowered and pulled in	Lids tightened	Lips pressed, jaw clenched
Scared	Inner brows raised	Lids wide open	Mouth stretched and open
Shocked	Raised	Lids wide open	Jaw dropped, mouth rounder than scared
Disgusted	Lowered	Lids narrowed	Like anger, with less lip pressure and fewer teeth showing

Treat these as starting points, then push the face toward the specific moment in the scene.

Aiming the Eyes With the View Target

On models built for SFM, the eyes do not need to be rotated one at a time. Under the Eyes group you will find a control called viewTarget. It is a point in the scene, and both eyes turn to look at it. Move the point and the gaze follows.

To aim the eyes, select viewTarget, switch to the move manipulator, and drag the target to where the character should look. Valve’s facial animation page describes selecting the view target, framing it with the work camera and selecting all of time in the Motion Editor with Ctrl+A when you want the change to hold for the whole shot. It also suggests moving the target’s pivot to sit between the eyes, which makes aiming easier to judge.

• Place the target on what the character looks at, not on the camera, unless the shot breaks the fourth wall.
• Keep the target at a believable distance. A target very close to the face makes the eyes cross.
• If the target vanishes, the Default procedural preset sends it back to the model’s root, usually at the feet, so you will need to move it up again.

Models without a view target may have eye sliders or eye bones instead. Eye bones show position and rotation values when selected. Rotate them with the manipulator or type exact values, and rotate both eyes together so they stay parallel.

Using Emotion Presets and Saving Your Own

When face controls are selected on an HWM character, the Emotion tab shows preset sliders for whole expressions. Each preset moves many control sliders at once. Because preset sliders blend and then reset, you can stack two presets, for example a touch of sadness on top of a smile, to get a mixed expression.

Once you have built a face you like, save it as your own preset so you can reuse it across shots. The Presets panel has a Create preset option for this. Community users report that it can be greyed out when a rig script is applied to the animation set, so create face presets before attaching a rig, or on a clean copy of the character. Presets can store bone values as well as flexes, which is how bone-based faces get reusable expressions. If you want to know how rigs change the animation set, our SFM rigging guide explains what a rig adds and hides.

Keyframing Facial Animation in the Motion Editor and Graph Editor

Faces are keyed with the same tools as the body. If either editor is new to you, start with our SFM Motion Editor and Graph Editor guide.

Hold, Change, Settle

Most facial acting follows a simple pattern. The face holds one expression, changes quickly when a thought changes, then settles into the new expression. In the Motion Editor, make a time selection where the change should happen, add falloffs so it eases in and out, and drag the sliders or a preset. In the Graph Editor, check the curves for each key slider and shape the tangents so the change has a fast middle and soft ends.

Avoid changing the whole face at once. Let the brows react first and the mouth a few frames later, so the change reads as a thought rather than a switch.

Blinks and Eye Darts

A face that never blinks looks dead, even with a perfect expression. Key the eyelid close slider so the lid drops quickly and opens a little slower. Place blinks at moments of change, such as just before a head turn or when a new thought arrives, rather than on a steady beat. During long holds, small quick moves of the view target add life.

Layering Expressions With Lip Sync

Dialog scenes need both speech and emotion on the same face. The cleanest order is to build the lip sync first, then add the expression around it. Extract Phonemes writes animation onto the mouth sliders, so mouth keys you set before running it can be replaced. Our SFM lip sync tutorial covers extraction and mouth cleanup in full.

Once the mouth reads well, split the face into layers of responsibility:

• Upper face for emotion. Brows, lids and eyes carry the feeling of the line, while phoneme extraction works mainly on the mouth.
• Mouth corners for mood. Push smile or frown controls on top of the speech shapes, using Offset mode in the Motion Editor over a time selection so the existing speech motion is kept.
• Jaw for energy. A slightly wider jaw on stressed words reads as louder or more emotional speech.
• Eyes for thought. Move the view target at the start of new sentences or ideas.

Check the result at full speed. A strong smile can stop the lips closing on sounds such as “m” and “b”.

Full SFM Face Posing Workflow, Step by Step

This walkthrough assumes one character with face flexes in a lit map and a shot camera already framed. Our SFM camera tips help you pick a framing close enough for the face to read.

When the face is finished, render a short test with our best SFM render settings.

Models Without Flexes: Bone-Based Faces

Some models, especially ports from other games, move the face with bones instead of flexes. You will see items such as jaw, eye or lip bones that show position and rotation values, but no face sliders.

• Pose face bones with the rotate and move manipulators in the viewport, or double-click a bone’s position or rotation entry to type exact values.
• Save useful bone poses as presets. Presets can store bone values, so a set of mouth shapes or expressions can be reused.
• Automatic lip sync needs phoneme presets. On a bone face you would have to build those presets yourself, one per phoneme.
• For real slider control, the face must be rebuilt with shape keys and recompiled, as our Blender to SFM guide explains.

Common SFM Facial Animation Problems

Problem	Likely cause	Fix
No face sliders appear	A bone is selected, or the model has no flexes	Select a face group, or pose face bones with manipulators
Only a few whole-face sliders	Standard model with combination sliders	Use the HWM version of the character if one exists
Emotion preset does nothing	Preset written for controls the model lacks	Build the expression from individual sliders
Expression looks too weak	Flex strength is capped at 1	Combine shapes, add jaw and brows, or edit the model’s flexes
L-R slider has no effect	The control is not stereo	Use a stereo control, or accept a symmetrical shape
Eyes point at the floor	View target reset to the root	Move viewTarget back in front of the face
Create preset is greyed out	A rig script is applied to the animation set	Create the preset before rigging or on an unrigged copy
Beard or mask ignores the face	Attached models have their own flexes	Key the matching sliders on the attachment as well

SFM Facial Animation FAQs