Composing expressions and character behaviour
Inspect fourteen jelly expression studies and separate body, gaze, lids, brow, mouth and timing before combining them.
Part 8 of 9From images to expressive 3D charactersSeries contents
- Building an expressive Giant Aperture mascot
- Turning reference pictures into a useful mesh
- Giving a mesh meaningful part labels
- Making painted eyes look and move like eyes
- Finding an ear and making it bend
- Turning a mouth line into an opening
- Fitting movement to the body in front of us
- Composing expressions and character behaviour
- Gaze, breathing and character behaviour
Curious is more than a raised eyebrow
Curious is a useful direction, but it is not a single movement. In the jelly study it combines a searching gaze, lifted tilted brow, a small uneven mouth and a slight upward reach. Separating those contributions lets us change the reading without rebuilding the character.
Named expressions provide a starting point. Direct controls provide the adjustments: shift attention, soften the lid, reduce the mouth opening or retain more body width. A saved expression should preserve the authored choices while leaving those contributions visible.
The map below is a snapshot of the actual fourteen-profile atlas. Open a study to inspect its target values and timing. It is a data view, with no 3D runtime or animation loaded. Values omitted from a profile are left to the shared defaults and layers; they should not be read as independently authored zeros.
Authored proposals · Requested values before the body response · Open a study to inspect it
CalmSlow easy blink; soft closed smile
Neutral; avoid unrelated breathing bounce.
- Requested body height
- 1.08 (rest 1.08)
- Upper lid
- 0.08
- Mouth curve
- 0.5
- Mouth width
- 0.5
Atlas timing
Entry: 0.12s start, 0.9s duration.
Blink: 1.25s start, 0.7s duration.
HappyLight blink, lifted brow and smiling corners
Small buoyant compression; smile does most of the work.
- Requested body height
- 1.0368 (rest 1.08)
- Upper lid
- 0.22
- Brow lift
- 0.12
- Mouth curve
- 0.95
- Mouth width
- 0.62
- Mouth opening
- 0.04
Atlas timing
Entry: 0.12s start, 0.9s duration.
Blink: 0.65s start, 0.35s duration.
DelightedEye squeeze and a small open smile
Brief stretch rebound, then smiling squash; distinguish from surprise.
- Requested body height
- 0.9936 (rest 1.08)
- Upper lid
- 0.65
- Brow lift
- 0.18
- Mouth curve
- 0.9
- Mouth width
- 0.6
- Mouth opening
- 0.3
- Eye aperture closure
- 0.24
Atlas timing
Entry: 0.12s start, 0.9s duration.
Body peak: 1.2096, centred at 0.62s, over 0.5s.
Blink: 1s start, 0.4s duration.
CuriousThought-change blink, lifted brow and searching gaze
Small upward reach; gaze leads.
- Requested body height
- 1.134 (rest 1.08)
- Upper lid
- 0.08
- Gaze left / right (degrees)
- 8
- Gaze up / down (degrees)
- 5
- Brow tilt
- 0.24
- Brow lift
- 0.18
- Mouth width
- 0.28
- Mouth curve
- 0.3
- Mouth tilt
- 0.25
- Mouth opening
- 0.04
Atlas timing
Entry: 0.12s start, 0.9s duration.
Blink: 0.45s start, 0.35s duration.
SurprisedFast widening; small round O; hold without a blink
Fast upward take, brief overshoot, hold a milder stretch. External shocks should not telegraph with a long anticipation.
- Requested body height
- 1.2744 (rest 1.08)
- Upper lid
- 0
- Brow lift
- 0.3
- Brow cap
- 0.94
- Mouth width
- 0.24
- Mouth curve
- 0
- Mouth opening
- 0.34
- Eye aperture closure
- 0.44
Atlas timing
Entry: 0.12s start, 0.24s duration.
Body peak: 1.3716, centred at 0.36s, over 0.36s.
No authored blink in this study.
WorriedHesitant glance and compressed mouth
Small protective compression; alert fear would be a separate stretch variant.
- Requested body height
- 1.0368 (rest 1.08)
- Upper lid
- 0.2
- Brow tilt
- -0.28
- Brow lift
- 0.1
- Mouth curve
- -0.35
- Mouth width
- 0.4
- Mouth opening
- 0.03
- Mouth tilt
- -0.1
- Gaze left / right (degrees)
- -5
Atlas timing
Entry: 0.12s start, 0.9s duration.
Blink: 0.55s start, 0.5s duration.
SadDownward gaze, slow blink and gentle frown
Slow downward settle; avoid comic rebound.
- Requested body height
- 0.972 (rest 1.08)
- Upper lid
- 0.44
- Brow tilt
- -0.18
- Brow lift
- -0.03
- Mouth curve
- -0.8
- Mouth width
- 0.44
- Mouth drop
- 0.14
- Gaze up / down (degrees)
- -5
Atlas timing
Entry: 0.12s start, 1.35s duration.
Blink: 1.1s start, 0.8s duration.
SleepyHeavy droop and a long closing blink
Deep slow sag; low energy and no bounce.
- Requested body height
- 0.9072 (rest 1.08)
- Upper lid
- 0.7
- Gaze up / down (degrees)
- -4
- Mouth curve
- 0.15
- Mouth width
- 0.3
- Mouth opening
- 0.13
Atlas timing
Entry: 0.12s start, 1.35s duration.
Blink: 1s start, 0.95s duration.
SkepticalHeld half-lid, tilted brow and a corner-led mouth
Almost neutral; brow and gaze carry the reading.
- Requested body height
- 1.0476 (rest 1.08)
- Upper lid
- 0.5
- Brow tilt
- 0.35
- Brow lift
- 0.05
- Mouth tilt
- 0.6
- Mouth asymmetry
- 0.4
- Mouth curve
- 0.05
- Mouth width
- 0.42
- Gaze left / right (degrees)
- -7
Atlas timing
Entry: 0.12s start, 0.9s duration.
Blink: 0.4s start, 0.3s duration.
FocusedSustained narrowing and small compressed mouth
Controlled scrunch; suitable preparation for a later ear pop.
- Requested body height
- 1.0044 (rest 1.08)
- Upper lid
- 0.38
- Brow tilt
- -0.1
- Brow lift
- -0.08
- Mouth width
- 0.32
- Mouth curve
- -0.05
- Gaze left / right (degrees)
- 2
Atlas timing
Entry: 0.12s start, 0.9s duration.
No authored blink in this study.
FrustratedRestrained compression, then a reset blink
Tight compression; optional brief release, avoid an angry tall silhouette.
- Requested body height
- 0.9612 (rest 1.08)
- Upper lid
- 0.48
- Brow tilt
- -0.22
- Brow lift
- -0.08
- Mouth curve
- -0.5
- Mouth width
- 0.43
- Mouth opening
- 0.07
Atlas timing
Entry: 0.12s start, 0.9s duration.
Blink: 1.75s start, 0.4s duration.
RelievedLong release blink and an exhale into a smile
Exhale from slight held tension into a soft squash.
- Requested body height
- 0.972 (rest 1.08)
- Upper lid
- 0.48
- Brow lift
- 0.08
- Mouth width
- 0.54
- Mouth curve
- 0.8
- Mouth opening
- 0.08
Atlas timing
Entry: 0.3s start, 1.1s duration.
Body peak: 1.1232, centred at 0.25s, over 0.5s.
Blink: 0.5s start, 0.75s duration.
SmugSettled compression, held half-lid, small asymmetric smile
Deliberate low broad settle; restrained asymmetry.
- Requested body height
- 0.972 (rest 1.08)
- Upper lid
- 0.48
- Brow tilt
- 0.18
- Brow lift
- 0.03
- Mouth curve
- 0.7
- Mouth width
- 0.42
- Mouth tilt
- 0.22
- Mouth asymmetry
- 0.15
Atlas timing
Entry: 0.12s start, 1.35s duration.
Blink: 0.5s start, 0.4s duration.
ContentSoft settled squash and relaxed centred smile
Soft symmetrical squash; centred relaxed smile.
- Requested body height
- 0.9936 (rest 1.08)
- Upper lid
- 0.38
- Brow lift
- 0.04
- Mouth curve
- 0.75
- Mouth width
- 0.48
Atlas timing
Entry: 0.12s start, 1.35s duration.
Blink: 0.7s start, 0.7s duration.
The atlas study returns after its hold. The live stage blends held targets and adds its own motion and ambient life; these timings are not a playback promise for its selector.
A held pose and a performance use the map differently
The atlas can sample a three-second sequence with entry, hold, blink and return. Some studies add a brief body peak. Surprise enters quickly and holds without an authored blink; Sleepy enters slowly and has a long closing blink. Timing changes what a related set of shapes communicates.
The interactive stage takes held expression targets and blends between them. It does not replay every authored atlas blink or peak whenever an expression is selected. The stage has its own gesture, gaze, body response and ambient blink layers.
This distinction prevents a misleading demo: a selector that reaches the right face has not necessarily reproduced the timed study. A comparison needs to identify whether it shows the held target, the authored sequence or the combined stage behaviour.
A new expression should inherit unfinished motion
An interactive character rarely gets to finish every movement before the next input arrives. If a second expression resets the first transition, the body can snap, pause or lose momentum at exactly the moment it should feel responsive.
The jelly’s damped transitions retain position and velocity when the target changes. The body has a separate response, including adjustable bounce and local flow. This lets a face change while the soft body is still catching up.
Gaze is layered in the same spirit. Optical aim responds before the eye placement and body follow. Filtering the incoming attention signal before resolving those layers gives a moving pointer a continuous influence instead of turning every sample into a fresh animation.
Use a blink to say something
A blink can mark a change of thought, release or attention. Shawn Kelly’s acting advice is a useful practitioner reference for treating it as a performance choice rather than an arbitrary repeating event.
The authored atlas makes those choices visible: Focused has no authored blink during its study hold, while Relieved has a long release blink. The shared ambient layer must still be considered in the combined result. A deliberate blink and an ambient blink should not be judged independently if they overlap on screen.
Material changes also need restraint. Warmth and clarity were useful concept variables, but the later acting studies focus on the face, attention and silhouette. An expression should remain readable without relying on a colour change to explain it.
Share the intent, respect the construction
The procedural jelly changes a generated surface. A reconstructed character may use bones, skin weights, fitted eyes and mouth shapes. A value that works for one mechanism is not a portable deformation recipe for the other.
For a fitted rig, the body’s supported controls and measured ranges establish what an expression can ask for. For the jelly, the body generator and silhouette response establish a different set of limits. The common goal is an editable direction whose face, body and timing can be understood.
The series starts and ends with that distinction: automate the mechanical work so a person can direct the character. A useful expression system makes the influential decisions easy to find, easy to change and possible to record.