Unreal plugin

The IAMX component

One component turns any actor into a connected digital human.

The IAMX Component is the heart of the plugin. Attach it to any actor — a MetaHuman, a stylized avatar, a robot, or an empty Pawn — and that actor gains the full IAMX pipeline: listening, understanding, speaking, lip sync, body animation, and memory. Everything is driven by a single connection ID you copy from the panel at iamx.live, so your level holds no API keys and no provider wiring. The component pulls its personality, voice, language, and enabled features from the cloud at runtime.

This page walks through adding the component, the settings you will touch most often, and how each interaction mode behaves. When you are done here, continue to animation nodes to drive the face and body, and start the conversation.

The component is self-contained. If you can drop an actor into a level and set one string, you have a talking, listening character. No graph wiring is required to get a first conversation running.

Add the component

The fastest path from an empty level to a working digital human is four steps. You will need a character connection ID, which you create in the panel (Quickstart covers making your first character if you have not yet).

  1. Drop your character into the level — drag a MetaHuman actor, your own avatar actor, or any Actor/Pawn from the Content Browser into the viewport. The component works with any actor; it only needs a mesh with morph targets if you want facial lip sync.
  2. Select it and Add Component → IAMX — with the actor selected, open the Details panel, click Add Component, and choose IAMX. The component appears in the actor's component tree and its settings expand in Details.
  3. Paste the character's Connection ID — copy the connection ID from your character's page in the panel and paste it into the component's Connection ID field. This binds the actor to that character's personality, voice, language, and feature set.
  4. Pick an interaction mode — set Interaction Mode to decide how conversations begin: push-to-talk, voice-activated, wake word, or proximity. Press Play and talk to your character.
The Connection ID is the only required field. Every other setting has a sensible default and can also be managed centrally from the panel, so a whole fleet of kiosks can be re-tuned without reopening the project.

What the component brings with it

Listen & understand

Captures speech from the microphone, cleans it, transcribes it, and streams it into the conversation with your chosen LLM — GPT, Claude, Gemini, or a local Ollama model.

Speak & sync

Generates the reply as natural speech and drives the character's mouth and face in real time, sentence by sentence, so audio and lips stay locked together.

Move & remember

Plays idle, talking, and listening body animation automatically, holds eye contact, and can remember returning visitors across sessions.

Key settings

These are the properties you will adjust most often in the Details panel. They live under the component's IAMX category.

PropertyWhat it doesDefault
Connection IDBinds the actor to a character in the panel. Supplies personality, voice, language, and enabled features. Required.
Interaction ModeHow conversations start: Push-to-talk, Voice-activated, Wake word, or Proximity. See the table below.Push-to-talk
Interaction KeyThe key held (push-to-talk) or pressed (manual) to talk. Rebind to any key.T
Proximity RadiusDistance, in world units, at which a Proximity-mode character wakes up and greets an approaching player.250
Show Chat UIToggles the built-in on-screen chat panel with a mic button, text field, and scrolling transcript. Handy for testing and for text-first kiosks.On
Leave Show Chat UI on while you build — it gives you a text input to talk to the character without a microphone, and a live transcript to confirm that speech recognition and the LLM are responding. Turn it off for a clean, voice-only kiosk.

Interaction Key

The interaction key does double duty depending on mode. In Push-to-talk the character listens only while the key is held and stops the moment you release it. In manual/testing flows a single press starts and stops a turn. The default is T; rebind it to a controller button, a spacebar, or any key that fits your rig or demo booth.

Interaction modes

The interaction mode decides how a turn begins. All four modes share the same understand-speak-sync pipeline underneath — they differ only in the trigger. Pick the one that matches your environment: a quiet desk demo, a noisy trade-show floor, a hands-free wall panel, or a walk-up kiosk.

ModeBehaviorBest for
Push-to-talkThe character listens only while the interaction key is held. Release to end the turn and get a reply. Zero false triggers because nothing is heard unless you are holding the key.Noisy rooms, demos, precise control.
Voice-activatedAlways listening. Detects when you start speaking and automatically stops on silence, then replies. No key required. An echo guard prevents the character from hearing its own voice.Hands-free, seated one-on-one sessions.
Wake wordIdles until it hears its name (or a phrase you configure), then opens a listening window for your request. Ignores everything else, so it can sit in an open space without reacting to passing chatter.Shared spaces, ambient assistants.
ProximityWakes and greets when a player enters Proximity Radius, and can wind down when they leave. Turns a walk-up into a conversation with no input at all.Kiosks, lobbies, exhibits.

Modes can be switched at runtime, and the panel can override the mode per deployment — so the same build can ship as push-to-talk on a desk and proximity on a kiosk. For the full turn lifecycle, interrupt/barge-in behavior, and cooldown tuning, see the conversation guide.

In Voice-activated and Wake word modes the microphone is open continuously. Test in the actual room: reflective surfaces and nearby speakers can feed the character its own audio. The built-in echo guard and post-playback cooldown handle most rooms, but a directional mic on a busy floor is worth it.

Choosing a mode

For a non-MetaHuman avatar — a stylized character, a purchased model, or your own rig — also enable custom character mapping on the component so IAMX drives your mesh's own blend shape names instead of the MetaHuman set. Full setup, including how to map visemes to your rig, is in custom avatars.

A minimal setup

You rarely need code, but if you prefer to add and configure the component from C++ or want to see exactly what the four steps do, it looks like this:

// In your actor's constructor or BeginPlay
UIAMXComponent* IAMX = CreateDefaultSubobject<UIAMXComponent>(TEXT("IAMX"));

// Bind to a character in the panel
IAMX->ConnectionID = TEXT("your-character-connection-id");

// How conversations begin
IAMX->InteractionMode   = EIAMXInteractionMode::Proximity;
IAMX->ProximityRadius   = 250.f;   // world units
IAMX->InteractionKey    = EKeys::T;
IAMX->bShowChatUI       = true;

// For a non-MetaHuman avatar:
// IAMX->bUseCustomCharacterMapping = true;

In the editor, the equivalent is: drop the actor, Add Component → IAMX, paste the Connection ID, choose a mode. Both routes end at the same place — a connected digital human ready to talk.

Verify it works

  1. Press Play — the character should idle naturally. If Show Chat UI is on, the chat panel appears in the corner.
  2. Trigger a turn — hold the interaction key, speak, and release (or approach the character in Proximity mode). Your words should appear in the transcript.
  3. Watch the reply — the character speaks, the mouth moves in sync, and body animation shifts to a talking pose. That confirms the whole pipeline — listen, understand, speak, sync — is live.
No audio but text is flowing? The pipeline is connected and the issue is playback or voice config — check the character's voice in the panel. No transcript at all? Confirm the Connection ID is pasted correctly and the microphone is selected in your OS.

Next

The component gets your character connected and conversational. From here:

Animation nodes →

Drive the face and body with the IAMX AnimGraph nodes: automatic lip sync, idle/talking/listening blends, and eye contact. See animation.

Start the conversation →

Kick off greetings, handle interrupts, wire actions, and control the flow from Blueprint. See start the conversation.

Related: Quickstart · Install · Conversation & modes · Custom avatars · Blueprint API