Timing & Blending¶
An emotion on a face is never simply on or off. It rises, holds for a while and falls away, and all the while other things are happening on the same face: another emotion arriving, the same one asked for again, the character talking.
This page covers each of those. Ramps is how one emotion moves over time. Overlapping Emotions is what happens where two meet on the face, whether they are different emotions or the same one twice. Emotion & Speech Share the Mouth is how an emotion and the words share one mouth.
Ramps¶
An emotion is not a switch. Each one is an envelope: where it starts, how quickly it comes up, how long it holds, how quickly it goes, and where it ends.
Those five points are what you shape, whether by dragging handles on the composer's emotion row or by typing exact times into its editor.
An Emotion's Own Ramps¶
Left alone, the ramps take the emotion's own shape. Every emotion ships with its own rise and fall, chosen for how that feeling moves: Surprised snaps on and is gone almost as fast, Angry builds and is slow to let go, and Sad creeps in and lingers longest of all. So an emotion you have not thought about still arrives and leaves like itself, not like a light switch.
They are used wherever a ramp is not set:
- A transcript tag without
rampUporrampDown, such as a plain<Angry>…</Angry>. See Transcript Tags. - A composer emotion cue with its ramps on Default.
- A baked emotion cue file played on a Timeline emotion clip set to Cue File.
- A script that plays an emotion without passing ramps, and the emotions a trigger palette fires.
The hold is not part of it. How long an emotion stays at full is always the span of the tag or cue that plays it.
A Timeline emotion clip set to Emotion takes them as its starting point. Its ease handles hold the rise and fall it plays, and a new clip is given these values to begin from. See Emotion on the Timeline page.
To watch an emotion's own ramps play on your character, open it in the Face Rig Designer's Emotion Preview, whose Rise and Fall start at exactly these values: Preview Timing.
Where the ramps sit relative to the words you tagged matters. An emotion that arrives exactly on its first word reads as late, because a real face is already changing before the sentence starts.
Overlapping Emotions¶
Two emotions at once resolve instead of fighting. Where they reach the same part of the face they combine, and the result is the same on every run.
That determinism matters more than it sounds. It means a line that looked right in the editor looks the same in the build, and the same on every playthrough. Nothing here is random.
Two Ways to Combine¶
FaceCue has several ways of combining two things that reach the same part of the face at once. Two of them are the ones worth choosing between, and they differ at one moment above all: the crossfade, where one is on its way out while the other is on its way in.
Contributed Max¶
The stronger of the two wins at each moment. Where both are partway through a crossfade, neither is at its level, so the face dips through the middle, the way two audio clips crossfading do.
Requested Max¶
The two are added together, but never past the strongest level either one asked for. Through a crossfade, the one arriving makes up what the one leaving gives up, so the face carries across without dipping, and it stops at the strongest request rather than climbing past it.
The other ways are for special cases, and the Emotion driver's inspector lists them beside these two.
Two Different Emotions¶
Two emotions on a face often reach different parts of it. Surprise is mostly in the brows and eyes, a smile mostly in the cheeks and the corners of the mouth. Wherever the two do not share anything, each plays exactly as it would on its own.
Where they do reach the same part of the face, the character's Emotion Blend decides, and it is Requested Max unless you change it. So two emotions that both raise the brows do not push them twice as high as either would alone, and when Worried fades out as Happy comes in, the face carries across from one to the other instead of sagging through the middle.
How worked up the character is follows the same setting, so two emotions that both carry agitation do not wind it up twice over either. See Mental State.
The Same Emotion Twice¶
An emotion asked for while it is already playing never doubles: a character asked to be happy twice is happy, not twice as happy. How its copies combine is a choice, made separately for two situations:
| Where the two copies come from | Setting on the Emotion driver | Default |
|---|---|---|
| Within one line, clip or set: two cues in one line's emotions, however they were written, two cues in one Timeline Cue File clip, or in one set played from code | Repeat Blend Within | Contributed Max |
| Between separate requests: a Timeline clip over a line that is already happy, two overlapping Happy clips on one track, or a script playing Happy while the line does | Repeat Blend Between | Contributed Max |
With the default, the stronger copy wins at each moment, so the face dips where two copies crossfade. How far depends on how they meet. Two tags side by side dip a little, while two composer cues with a short gap between them dip plainly. For copies that should read as one stretch, set the matching setting to Requested Max, which carries the face across as a whole. A single part of the face can still dip for a moment, where one copy's version of it has already gone before the next copy's arrives.
A demeanor is the exception. Each demeanor palette carries a Repeat Blend of its own, which starts on Requested Max, so a demeanor drawing the same emotion twice in a row carries straight across. See Demeanors.
Overlapping Tags¶
Tags are settled one step earlier, when the line is baked. Two tags of the same emotion whose words overlap become a single cue only when they would bake to the same thing: the same intensity, the same ramps, the same ramp style. If anything differs between them, both are written, and the line plays them as two copies of one emotion, meeting the way Repeat Blend Within says.
A peak inside a milder stretch. A mild Happy across the sentence, with a stronger one around a word in the middle:
<Happy i=0.4>It was so <Happy i=1>good</Happy> to see you</Happy>
The two ask for different strengths, so both are kept. The sentence opens mild, lifts to the peak on good, and settles back to mild for the rest of the line. The mild Happy is still running underneath the peak the whole time, so the face comes back down to it rather than through it.
<Happy i=0.4>, with <Happy i=1> around a word inside it. The thin lines are the two cues on their own, the thick line is the face.The other way round. Put the stronger tag on the outside and the milder one within it, and the stronger covers it. The face plays the strong Happy across the whole sentence, because a second copy of an emotion never takes away from what is already on the face.
Two tags that ask for exactly the same thing are a different matter. They have nothing to tell apart, so they become one cue covering the words of both, which is what you want when a sentence is tagged twice over for the same feeling.
Tags that cross. Give the tags an id and they can share a stretch of words rather than sit one inside the other. Say the mild Happy is meant for It was so good and the strong one for good to see you, overlapping on the word between them:
<Happy id=a i=0.4>It was so <Happy id=b i=1>good</Happy id=a> to see you</Happy id=b>
Each closing tag names the opening tag it closes, so the mild one ends on good while the strong one runs to the end of the sentence. The face opens mild, rises on good, and stays up from there, because the strong cue is still going after the mild one has finished.
An id is yours to name, it works on every kind of tag, and you only need one where tags of the same name overlap. If a closing tag names an id that no open tag is using, or two open tags of one emotion are given the same id, the baker says so and pairs them the way it would without the id.
Leave the ids out and the tags nest. The same words without them read as one tag inside the other, because a closing tag with no id goes with the nearest opening tag that has none:
<Happy i=0.4>It was so <Happy i=1>good</Happy> to see you</Happy>
The first </Happy> closes the strong tag around good alone, and the mild one runs on to the end. That is the peak-inside-a-milder-stretch reading from the top of this section, and it is what you get whenever same-emotion tags overlap without ids.
The mood a line leaves behind comes from the cue that finishes last. Where a peak sits inside a milder stretch, that is the milder one, so the character carries the gentler feeling on rather than the height of the peak.
Choosing How They Meet¶
The character's own settings, the three above, decide how emotions combine, and they suit most scenes as they ship. One moment can ask for something different. A Timeline emotion clip carries its own Emotion Blend and Repeat Blend, and code can pass the same choice when it plays an emotion. A choice made there covers every copy of that clip's or that call's emotions, whether they meet each other or something else. See Emotion on the Timeline page and From Your Own Code.
Tags and cues placed in the Cue Clip Composer cannot ask for a choice of their own yet, so the emotions on a line always meet the way the character's settings say.
Emotion & Speech Share the Mouth¶
Speech and emotion play on the same face at once, and FaceCue shares it between them so neither fights the other.
The brows and eyes stay with the emotion. Talking does not take them away.
The cheeks carry both. The cheek half of a smile holds whatever the mouth is doing.
On the mouth, the words come first. Where speech and an emotion want the same mouth blendshape, the words lead so they stay readable, and the emotion keeps a share of it. A mouth blendshape speech is not using keeps the emotion in full. If an emotion with a strong mouth still crowds the words, set Lower Scope on the FaceCue Resolver to Per Region, and the emotion gives way across the whole mouth while the character speaks. See Emotion and Speech on the Mouth.
When the lips have to close for a word, emotion gives the mouth up for that moment. But the closure only takes what it needs, so the expression bends around it and comes back instead of snapping off. A smiling character still looks like they are smiling through the word.
This Is Why a Smile Survives a Sentence
The naive version of this is a priority list, where speech wins the mouth and the smile disappears every time a word closes. The face flickers, and it is one of the most common tells in facial animation.
Sharing the face instead means the parts that can coexist do, and only the part that genuinely conflicts yields.