Speech

The part that forms the mouth. Which source it follows, how the shapes run into each other, and the rules that keep speech readable when blending alone would smooth it away.

This is the driver most projects spend their time with, and the one the rest of the performance is layered over.

Drive Modes

Where the mouth's instructions come from. One character is in one mode at a time, chosen under Mode on the Speech driver, and the inspector titles its settings by the same names.

Mode	What drives the mouth
Baked Phoneme	The cues from a bake. The most accurate, and the one to use for dialogue you have
Live Viseme	Live recognition, from a microphone or from a clip with no bake behind it
Live Amplitude	The loudness of the audio, with no recognition at all
Disabled	Nothing. The mouth is left to whatever else is driving it

Baked Phoneme is the one to build around. It knows which phoneme is happening and when, so everything downstream, the rules, the delivery, the timing, has something exact to work from.

Live Viseme is for speech that does not exist ahead of time: a player's microphone, voice chat, a line generated while the game runs.

Live Amplitude is the simplest. It opens the mouth in proportion to how loud the audio is, and it needs almost nothing: no bake, no model, only the rig map the wizard already filled in. It also runs wherever audio can be read at all, which includes the places Live Viseme cannot go, such as a browser build. That makes it the usual choice for live audio when you want the least setup, and the usual thing to fall back on when Live Viseme is not available.

Baked & Live

A baked line knows what is coming. A live line is heard as it plays. That one difference decides most of what the mouth can do, so the rest of this chapter is split in two.

	Baked Phoneme	Live Viseme	Live Amplitude
Shapes run into each other	Yes	Yes	No, loudness bands instead
Mouth sets up ahead of the voice	Yes	No	No
Delivery styles	Yes	No	No
Lips close where a word needs it	Yes	Yes	No
Reads which phonemes were stressed	Yes	No	No
Base pose	Yes	Yes	No
Plays without audio	Yes	No	No

• Baked covers Baked Phoneme: coarticulation, anticipation, speaker delivery, the rules underneath, loudness and prominence, the base pose, laughs, and playing a line with no audio.
• Live covers Live Viseme and Live Amplitude: what each needs, what each keeps from a baked line, and how to choose between them.

The Fallback Chain

A mode you asked for is not always a mode that can run. A clip might have no bake, or a character might be speaking audio that arrived a second ago.

Instead of stopping, FaceCue falls through a chain, in this order: Baked Phoneme, then Live Viseme, then Live Amplitude. A clip with no cues is read live if live recognition can run, and by its loudness if not, so the mouth is roughly right instead of doing nothing while the audio plays. Any mode that cannot run is passed over the same way, whether it lacks a bake, a rig map, audio it can read, or a platform it runs on. The chain is a list under Mode, Fallback Chain, and you can reorder it or take a tier out.

Every time a line plays in a mode other than the one you chose, the Console says so: Speech Fallback Engaged names the mode that took over and why the ones before it could not run.

Rough Means Fallback, Still Means Check the Console

If a line plays and the mouth looks rough or muddy, a fallback has almost certainly happened: the character is a tier or two below the mode you meant, usually because the clip's cues were not found. The Console names it. Start at When a Clip Will Not Load.

If the mouth does not move at all, look for Speech Fallback FAILURE. It means no mode in the chain could run, and it gives the reason for each. If nothing is logged, the audio never reached FaceCue. Check that the speaker's audio source is the one FaceCue is watching before looking anywhere else.