Getting started

How it works

Design a character in the panel, connect it to conversation services and present it in a browser or Unreal application.

The cloud workflow has three parts. You design the character in the panel. IAMX and the configured providers handle its conversation, including model responses, voice generation and facial animation. A client presents the result: a browser-rendered avatar, a streamed character or your Unreal application. Rendering, microphone input and playback still depend on the visitor's device and chosen presentation.

This page explains the moving parts, how they talk to each other, and the two runtime modes you can choose from. If you just want to get a character talking, jump to the Quickstart; if you need to install the plugin, see Install.

IAMX dashboard showing characters, monthly interactions, knowledge store and plan — The dashboard: a live view of what the workspace is doing.

The three parts

These parts have different responsibilities. Knowing which one is involved helps you find the right setting when something needs to change.

1. The panel

Open the IAMX panel to create characters, write a persona, add knowledge, choose a voice, configure actions and inspect conversations. Avatar Studio controls the saved appearance and scene; Experience adds website guidance and other interactions.

2. The cloud

IAMX coordinates incoming messages, the selected AI provider, available knowledge and tools, and the chosen voice. The standard speech path also prepares facial animation from the generated audio. The exact sequence depends on whether the character uses standard or realtime voice.

3. The client

The browser or Unreal application handles interaction and playback. Browser-rendered GLB avatars use the visitor's graphics hardware; Pixel Streaming displays video rendered on a server. Supported web conversations can use the browser's speech recognition before sending a message.

What the panel owns

The panel is the single source of truth for a character's identity and capabilities:

Persona & instructions — who the character is, tone, language, guardrails.
Voice — the voice the character speaks with, including custom and cloned voices.
Knowledge — documents and facts the character can draw on when answering.
Actions & tools — the things the character is allowed to do (call an API, trigger an animation, hand off to a workflow).
Behavior tuning — greeting logic, interaction mode, memory, and per-feature settings.
Reports — transcripts, summaries and usage per person and per character.

What the cloud owns

The cloud coordinates a session, checks access and usage, passes messages to the selected model and returns its response. Standard voice generation and facial animation are separate services. Tools and knowledge retrieval can add steps to a reply. Browser speech recognition and native realtime voice take different input paths; see Voice and microphone.

What the client owns

The client handles microphone controls, the live connection, audio playback and the selected visual presentation. A GLB needs compatible facial shapes and body animations; an Unreal character needs the appropriate rig integration. Scene actions and media also have their own playback controls. See Avatar Studio for web setup or Blueprint API for Unreal events and nodes.

The connection

Each character has a unique Character ID. Use it in the Unreal component's connection settings or copy the website widget code from Publish, which includes data-character. The ID selects the character; visibility and authentication determine who can open a conversation. An ID alone does not grant access to a private character.

Once connected, the client sends text or supported audio input and receives response text, audio and presentation events. Playback can begin before the full reply has finished on supported streaming paths. It still takes time to detect the end of speech, generate a response and prepare audio; model, provider, tools and network conditions affect the delay.

  ┌─────────────┐         ┌──────────────────────┐         ┌──────────────┐
  │   PANEL     │         │     IAMX  CLOUD      │         │    CLIENT    │
  │  (design)   │         │   (the character's   │         │  (your app / │
  │             │         │       brain)         │         │   embed)     │
  └──────┬──────┘         └──────────┬───────────┘         └──────┬───────┘
         │                           │                            │
         │  persona · voice ·        │                            │
         │  knowledge · actions      │                            │
         ▼                           │                            │
   ┌───────────┐                     │                            │
   │ connection│─────── issued ──────┼───────────────────────────▶│  paste ID
   │    ID     │                     │                            │  into client
   └───────────┘                     │                            │
                                     │        connect (ID)        │
                                     │◀───────────────────────────│
                                     │                            │
                     mic audio / text│  ▲  ▲  ▲                   │  ── UP ──▶
                                     │  │  │  │                   │
                    voice + face +   │  ▼  ▼  ▼                   │  ◀─ DOWN ─
                    emotion + gesture│  (streamed frame by frame) │
                                     │                            ▼
                                                            renders on
                                                          MetaHuman / avatar

Use the connection details generated in Publish for your chosen channel. Keep private credentials out of public website code; use the public widget workflow for a character intended to serve website visitors.

A session, start to finish

Design — you build the character in the panel and it receives a connection ID.
Connect — the client opens a session for that character, subject to access and account availability.
Speak — the visitor types or uses the microphone. The selected input path supplies text or audio.
Think — the cloud understands, reasons, and composes a reply, pulling on the character's persona, knowledge and tools.
Perform — the response is spoken using the selected voice and paired with the supported facial animation and action cues.
Present — the client plays the response and runs the configured visual experience.

Why a thin client

Putting the intelligence in the cloud and keeping the client thin is a deliberate design choice, and it pays off in three concrete ways.

Server-side conversation

The cloud conversation services do not require you to bundle their AI models into the visitor's application. Avatar assets, WebGL rendering or Unreal rendering still have their own memory and graphics requirements.

Update character settings

Change supported character settings in the panel without rebuilding the app. Save and publish the relevant configuration, then start a new session. New client features or plugin changes may still require an application update.

Choose a presentation

Use a supported GLB, LiveFace or Pixel Streaming presentation on the web, or integrate the Unreal plugin. Custom avatars need compatible facial mapping and animation; changing the model can require additional setup.

Start with the presentation your audience needs, then test it on their actual devices. A successful desktop conversation alone does not verify mobile microphone permission, frame rate or a website's embed policy.

Realtime vs standard

Core AI offers standard models and supported realtime speech modes. The main difference is whether a separate voice provider speaks the generated text or the selected model produces audio itself.

Realtime mode

Native model audio. The selected realtime service handles speech input and audio output, using its own voice list. It supports a different conversation path from a standard text model with separate TTS. Test interruptions, knowledge requirements and actual response time for your use case.

Standard mode

Separate model and voice. The language model produces a reply and the selected TTS provider speaks it. Supported paths can begin playing portions of the reply before the whole turn finishes. Voice choices, including any cloned voices, depend on that provider and your account access.

	Realtime	Standard
Response time	Depends on the realtime service, input and network	Also includes the separate model and speech-generation steps
Voice choices	Native realtime voices	Voices supported by the selected TTS provider
Interruptions	Test the selected realtime mode and client	Test the selected microphone mode and client
Best for	Live kiosks, fast dialog, demos	Brand voices, narration, wide language range
Set where	Character settings in the panel

Mode and model are both chosen in the panel, and you can change them any time. For how to pick the reasoning model behind your character — GPT, Claude, Gemini, or a local Ollama model — see Models.

Mode is a per-character setting, so you can run a snappy realtime greeter and a flexible-voice narrator side by side, each with its own connection ID.

Save, test and publish

Save character edits and test them in Model & Face → Test. Use Publish to apply the saved Studio scene to public widgets and embeds. Website guidance has separate Save experience and Publish experience controls. Open a new visitor session to check the published result.

See Avatar Studio for appearance, scenes and animation, Website embed for installation and Website experience for page awareness and navigation.

Next steps

Quickstart — create a character, grab its connection ID, and get it talking.
Install — add the plugin to your Unreal project. For a website, use Website embed.
Models — choose the reasoning model and runtime mode.
Blueprint API — the events, nodes and hooks the client exposes.
See it live at iamx.live, get the plugin on GitHub (releases), or ask on the forum.