Features

Lip sync & face

The full facial performance follows every word — streamed frame by frame.

When your character speaks, IAMX drives the entire face in real time. This isn't a mouth flapping open and shut on a volume meter — it's a full performance: jaw, lips, cheeks, and brows all moving together, plus eyes that find the player and a head that turns to meet them. Everything arrives as a live stream while the audio plays, and every layer lands as an AnimGraph node that blends on top of whatever your character is already doing.

New here? Get a character talking end-to-end in a few minutes with the Quickstart, then come back to tune the face.

Full-face performance

Speech is a whole-face event. A person doesn't just open their mouth to make an "ah" — their jaw drops, their lips round, their cheeks tighten on plosives, and their brows lift on an emphasized word. IAMX reproduces all of it. As each sentence is spoken, the facial performance is generated in the cloud and streamed back to the engine one frame at a time, staying locked to the audio you hear.

Jaw & lips

Full articulation — open vowels, closed bilabials, rounded and spread shapes — so the mouth reads as real speech, not a loop.

Cheeks

Subtle tension and puff around plosives and emphasis give the lower face the weight that sells a spoken line.

Brows

Upper-face motion tracks emphasis and phrasing, so questions, stresses, and pauses land on the face, not just in the audio.

Streamed, not pre-baked

Because the performance streams as the sentence plays, there's no bake step, no per-line asset to manage, and no lag between "the audio started" and "the face moved." The first frames arrive as speech begins and continue frame by frame until the line ends. Combined with sentence-level audio queuing, the character can start speaking the opening of a long reply while the rest is still being produced — the face stays in step the entire time.

Behaviour	What you get
Motion source	Full-face performance (jaw, lips, cheeks, brows), not amplitude-driven mouth open/close
Delivery	Streamed frame by frame in real time, synchronized to the spoken audio
Latency	Face begins moving as the line begins — no pre-render or bake
Long replies	Speech starts on the first ready sentence; the face follows each sentence as it plays

The performance is applied directly to the character's facial controls, so it reads correctly regardless of how the rest of the body is posed or moving.

Eyes & head

Eye contact is what turns a talking model into someone who is present. IAMX gives the character a real gaze: the eyes find the player's camera and hold it, and the head turns toward the player with natural limits and smoothing — the way a person keeps you in view without snapping their head around or staring through you.

How the gaze behaves

Find & hold — the eyes lock onto the player's camera and track it as it moves, so the character keeps looking at whoever they're talking to.
Head follow — the head turns to face the player, but only within a comfortable range; extreme angles are handled by the eyes rather than an unnatural neck twist.
Natural limits — yaw and pitch are clamped so the character never over-rotates. When the target moves past the limit, the head eases toward it instead of pinning at the edge.
Smoothing — motion is damped so the gaze glides to a new target rather than teleporting, which reads as attention rather than a tracking script.

The result is a character that makes eye contact like a person: attentive while listening, engaged while speaking, and never uncomfortably locked or robotically snapping between targets.

Eye contact pairs naturally with conversation state. While the character is listening it holds the player's gaze; the same drivers keep the eyes on target through the reply, so the whole exchange feels like a two-way conversation.

Non-destructive

Lip sync, eyes, and head all run as AnimGraph nodes that blend on top of whatever animation is already playing. Your idle, your gesture animations, your own state machines — none of it has to be replaced or rebuilt. The face layers over your existing body performance instead of fighting it.

Your existing AnimGraph
┌─────────────────────────────────────────────┐
│  Idle / gestures / your own state machines   │  ← base pose (unchanged)
│                    │                          │
│                    ▼                          │
│        ┌────────────────────────┐             │
│        │  IAMX Lip Sync (face)  │  ── blends on top
│        └────────────────────────┘             │
│                    │                          │
│                    ▼                          │
│        ┌────────────────────────┐             │
│        │  IAMX Eyes  +  Head     │  ── layered gaze
│        └────────────────────────┘             │
│                    │                          │
│                    ▼                          │
│                Final pose                     │
└─────────────────────────────────────────────┘

Because each layer is additive, the character can walk, gesture, and emote from your own animation setup while IAMX drives the face and gaze. Nothing about your body pipeline changes — you drop the nodes in and they take over only the channels they own. See Animation for how to slot the nodes into your graph and order the layers.

Because the face nodes blend additively over the base pose, put them after your body state machine in the graph. If a downstream node overwrites the same facial channels, it will win — layer order decides who has the final say.

Any avatar

The facial performance isn't tied to one character format. IAMX drives the standard MetaHuman facial controls out of the box, and it can just as easily drive your own custom character's blend shapes — you provide the mapping from the performance channels to the shape names your rig uses.

Character type	What you do
MetaHuman	Works out of the box — the standard facial controls are driven directly, no mapping required.
Custom avatar	Provide a mapping from IAMX's performance channels to your rig's blend shape names, and the same full-face performance drives your character.

This means a face you sculpted, bought, or imported can carry the exact same streamed performance as a MetaHuman — full jaw, lips, cheeks, and brow motion — as long as your model exposes blend shapes to target. The full walkthrough, including how to build and verify your channel map, is in Custom avatars.

Pick your character — a MetaHuman needs no setup; a custom avatar needs blend shapes on its facial mesh.
Map the channels — for a custom avatar, point each performance channel at the matching blend shape name on your rig.
Drop in the nodes — add the IAMX face and gaze nodes to your AnimGraph as described in Animation.
Speak and verify — trigger a line and watch the full face move; adjust the map if any shape reads wrong.

Tuning

Responsiveness is a taste decision, so IAMX exposes it. Two speeds shape how the face reacts: how quickly motion smooths toward each new target, and how quickly it fades as speech settles or ends. Dial them to make a character feel crisp and reactive, or soft and relaxed.

Control	Lower value	Higher value
Smoothing speed	Softer, more gradual mouth and face motion — gentler, calmer read	Snappier, more immediate reaction that tracks fast speech tightly
Fade speed	Shapes ease out slowly for a smoother settle between words and lines	Shapes release quickly, giving a sharper, more clipped delivery

The same philosophy applies to the gaze: the eye and head smoothing that keeps eye contact from snapping is adjustable, so you can make the character glance and settle gently or track the player more assertively.

Pro tip: start from the defaults, then nudge one value at a time while a character is actually speaking. Small changes to smoothing read very differently at conversational pace than they do on a static preview — always tune against live speech.

Where to go next

Wire up the graph

Layer the face and gaze nodes over your body animation in Animation.

Bring your own model

Map the performance onto a custom rig in Custom avatars.

See it live

Watch a character make eye contact and speak on the live demo.

Questions about a specific rig or a tuning combination that isn't behaving? Ask on the community forum, read build notes on the blog, or grab the latest build from releases.