Preview & Scrubbing

Watching a performance without entering play mode.

Using the Cue Clip Composer as an editor-time previewer: play a performance on a character without entering play mode, scrub it, and choose which parts of the face are moving while you watch.

Assign a character in Brain, turn Preview on, and press Play. The character speaks in the scene view, at edit time, with no build and no play mode. This is where most tuning happens, because the loop from changing something to seeing it is a few seconds long.

Choosing What Moves

The Preview dropdown decides which drivers move the character while preview is on. It covers Speech, Prosody, Emotion, Eyes and Breath, and each is independent.

Turning things off is the point of it. A face doing everything at once is hard to judge, so watching the mouth on its own tells you whether the mouth is right, and putting emotion back afterwards tells you whether the two work together.

Why an Entry Might Be Greyed Out

An entry is dimmed when the character has no such driver on it. A character with no breath driver has nothing to switch on, so the entry says so instead of pretending.

Breath has one extra condition. Its driver also has to be set to its active mode, because a breath driver that is not breathing has nothing to show either.

The Character Is Always Speaking

While preview is on, the Eyes entry reads Eyes (Talking).

That is the window telling you which state it is holding. Eyes behave differently in conversation than at rest, and preview asserts a speaking character throughout, so what you are watching is the talking behaviour rather than an idle one. The label changes back when preview is off.

It matters because the eyes you see here are the eyes of a character mid-line, which is the version worth judging when you are working on a line.

Play, Pause & Scrub

Play runs the clip forward with audio, and the character performs it.

Dragging the playhead scrubs, and the face follows to wherever you put it. Single-clicking a cue in the Visemes row scrubs to the start of that cue, which is the quickest way to land on a particular phoneme.

Playing it forward is the one to judge by. Time is really passing, so everything that depends on time behaves the way it will in your game.

Scrubbing and pausing are close, not identical. A held frame has no time passing in it, so the smoothing that softens the face in motion has nothing to work across, and you see the pose without it. The difference is usually small, and it tends to make a paused frame look a little sharper than the same moment does in motion.

Breath is the exception, because it is a cycle rather than a pose, and a frozen one would mislead. Wherever you land, the breath is where it would be at that moment of the line.

Judge a Line by Playing It

Use scrubbing to find a moment and to check what the cues are doing at it. Use playback to decide whether the line is right, because smoothing and settling are part of how it will look and they only happen while time is moving.

If a paused pose looks harsher than you expected, play the clip before changing anything.

Previewing in Isolation

Two habits are worth having.

Watch the mouth alone first. Everything else is layered on top of it, so a problem in the mouth reads as a problem everywhere. With Emotion and Eyes off, what is left is the line itself.

Then put one thing back at a time. A face that looks wrong with everything running is hard to diagnose, and the driver that is causing it is usually obvious the moment you switch it off.

The Sweep

Some rows are read out of the cue files. The rest are worked out by playing the clip, so the window has to play it before it can draw them.

Sweep is that run. It plays the whole performance through and fills in everything that is computed instead of stored.

That split does not follow the tiers. Final Output is computed, so the row that tells you what actually lands on the face needs a sweep as much as the deep rows do, even though it is there at Basic. What does not need one is the material read straight from the files: the waveform, the words and letters, the cues themselves, and the two authored timelines.

Auto keeps it current. With it on, FaceCue re-sweeps by itself when something that would change the result changes, on the interval beside it, so tuning a setting and looking at the rows does not need a button in between. Leaving it on is the reason most people never think about sweeping at all.

Export writes out what the sweep produced, so it wants one too.

An Empty Computed Row Is Telling You Something

Sweep results do not survive a script recompile, so the computed rows go blank after one and come back on the next sweep. With Auto on, that happens without you noticing.

If they stay blank, the sweep is not failing to have run, it is unable to run, and FaceCue says which of the two reasons it is: no character assigned, or a character whose mouth shapes have not been mapped yet.

Who Drives the Face

Several things can write to a character's face, and FaceCue never lets two of them do it at once. The face goes to whichever has the stronger claim, and the one that gives way is reset cleanly, so nothing is ever left half-driven. The order, from strongest to weakest:

1. The live performance. In Play Mode the character's own drivers own the face, and both editor tools stand aside. The Cue Clip Composer stops its preview and will not start one again until you are back in Edit Mode, where it picks preview up on its own. The Face Rig Designer locks, and the lock can be pressed to edit anyway for that play session, returning the next time you enter Play Mode.
2. The Cue Clip Composer, playing or scrubbing. A performance being played beats a recipe being tuned.
3. The Face Rig Designer, driving. With Drive or Mirror on, a recipe selected, the blendshape picker open or a preview running, the designer holds the face.
4. The composer, previewing but idle. A composer with Preview on and nothing playing only wants the face, and gives it up to a designer that is driving.

So the two editor tools divide the work naturally. The Face Rig Designer previews one recipe at a time, and the Cue Clip Composer previews a whole performance, a baked line with everything running, which is where to go once the recipes are shaped, to see them in a sentence. Press Play in the composer and the designer steps back. Stop, and the designer picks the face up again.

The Tuning Loop

The reason all of this is in one window is the loop it makes.

Change a recipe in the Face Rig Designer or a setting on the character, come back to the composer, and the change is already in what you are looking at. Play the line, judge it, change it again. No build, no play mode, and no waiting for a bake, because none of this re-bakes anything. The cues are what they were. What changes is what the character does with them.