Character: channels, faces & publishing
How you put the character in front of people, how it is drawn, and whether it recognises the person in front of it.
All of this lives in the character editor. Open the character from Characters (Open studio) and use the setting groups in the right dock:
| Group › tab | What it is for |
|---|---|
| Channels › Publish | Visibility, publishing, the website widget and every connection snippet. |
| Channels › Pixel Streaming | Show the character as a cloud MetaHuman. |
| Channels › Phone | The character's phone number and greeting, plus a test dialler. |
| Channels › WhatsApp | How the character answers on your WhatsApp Business number. |
| Channels › Publish preview | Rehearse the real website embed before publishing. |
| Channels › Versions & publishing | The scene draft, publishing a named release and the published history. |
| Channels › Capture & export | Scene variations, brand look, scene export and import, and image capture. |
| Identity & model › LiveFace 2D | A talking head made from one portrait photo. |
| Behavior › FaceID | Face recognition and ID-card verification for kiosks. |
A tab with a padlock is not included in your plan; selecting it opens Plan & Quota. LiveFace 2D and FaceID need the Pro plan or higher in the standard plans; Phone and WhatsApp are Enterprise features.
Publish — putting it in front of people
The Publish Status card shows what the character is currently (PRIVATE, UNLISTED or PUBLIC). Changing Visibility saves immediately.
| Visibility | Who can talk to it |
|---|---|
| Private — only you | Authenticated access is required. Test it with Test conversation while building; a public visitor cannot open a private embed without a key, and Copy website code stays disabled. |
| Unlisted — anyone with the link | Anyone with the link. The usual choice for a website widget or embed. |
| Public — everyone | Everyone. |
- Save your work. Finish the look, voice and scene changes and press Save in the toolbar (or Ctrl/Cmd+S). Publishing also saves pending edits and stops if they cannot be saved.
- Choose Publish. Use Publish in the editor toolbar, or Publish / Republish in the Publish Status card. This includes the saved Studio background, lighting, camera and character placement. The toolbar state Saved · Scene not published means the saved scene is still waiting to be applied to widgets and embeds.
- Open published scene. Check the full scene, voice and animations in a new visitor session.
- Choose a website template. In the Website Templates card, preview its size and opening state, then choose Copy website code. Follow Website avatars & embed for installation.
The transparent Floating Companion template hides scenery intentionally. To show the background, choose a scene or portrait template and make sure the Studio scene is published.
For page awareness, spoken website help and navigation actions, also use Save experience → Publish experience on Scene & motion › Experience; see Website experience. Publishing the character does not publish the Experience.
The Publish tab also provides the connection details for each channel:
- Character ID — the identifier the Unreal plugin, the web SDK and a direct WebSocket connection need.
- UE Plugin (Cloud Mode) — the settings block to paste into the IAMX component, including where your API key goes.
- WebSocket (JavaScript) — the starting point for your own app: open the socket with an API key and send
start_sessionwith the character ID. - HTML Embed (iframe) — a one-line iframe for any page, with no API key needed for an Unlisted or Public character. When the saved scene is newer than the published one, the card offers Publish scene.
API keys are created under My workspace › API Keys.
Publish preview, versions and capture
Three scene tools sit next to Publish in the Channels group:
- Publish preview runs the actual website embed inside the studio so you can inspect it before going live. Preview unsaved scene sends unsaved scene changes only to this test. Tests can produce real conversations.
- Versions & publishing keeps a cloud draft, compares it with earlier versions, publishes it under a Release name (Publish scene) and lists the Published history (the latest 20 versions). Restoring a version changes your draft; publish it to make it live.
- Capture & export keeps up to 12 Scene variations, copies and applies a brand look, exports and imports a scene file, and saves an image of the scene.
Pixel Streaming — a cloud MetaHuman
This gives the character the body and face of a full Unreal MetaHuman, rendered on a GPU server and streamed to the browser as video. Your visitors need nothing but a web page. You only choose the look — the voice, the knowledge and the rules all stay as you configured them on the other tabs.
On Channels › Pixel Streaming, click an avatar card to assign it, and Disable to go back to the character's own look. If no cards appear, no MetaHuman avatars have been added to the platform yet. Rendering infrastructure is separate from your conversation plan; see Other surfaces.
LiveFace 2D — a talking head from one photo
LiveFace renders the character as a real photograph whose mouth moves with the speech, instead of a 3D model. It is the fastest way to a convincing face: one good headshot and you have a talking presenter. You find it under Identity & model › LiveFace 2D (also reachable from the LiveFace 2D button in the character packs bar on the stage). The tab shows the flow Portrait photo → Lip sync → Embed.
- Upload a portrait. If the upload area reads Select IAMX Render Engine to upload a portrait, first choose Use IAMX Render Engine in the IAMX Render Engine section. Then, under Avatar (face), choose Upload portrait: a well-lit head-and-shoulders photo facing the camera, JPG or PNG up to 8 MB. A photo is required before the render engine can start.
- Switch the render target on. In Render target, turn the switch to On (it becomes available once a portrait is saved). When it is on, embeds show the 2D talking head instead of the 3D avatar. There is one engine; no engine choice is needed.
- Press Live preview under Preview & Embed to check that the mouth moves. The engine starts on demand; the first connection may take a few minutes.
- Put it on your site. For an Unlisted or Public character, use the website code (Copy website code) or the HTML Embed (iframe) on Channels › Publish: they need no API key and show the 2D face while the render target is on. Copy embed code on this tab gives an iframe with an
IAMX_API_KEYplaceholder for a key from My workspace › API Keys (Create API key takes you there); use it only for a Private character on a page you control, because the key is visible in the page source.
FaceID — recognising the person
With a camera attached, the character can tell whether the person in front of it is someone it has met before, and greet them accordingly. Where identity really matters — banking, for example — it can also match the live face against the chip of a Turkish ID card read over NFC. The tab is Behavior › FaceID, titled FaceID & Identity, with the flow Camera → Recognize → NFC ID → Greet / verify. Switch it on with Face recognition + identity verification.
Server, matching and timing
| Setting | Notes |
|---|---|
| FaceID server | Your own FaceID server, or empty for the platform's shared FaceID. |
| Match threshold (0–1) | Higher is stricter. 0.40 is recommended. |
| Exit delay (s) | How long after the person leaves the camera the conversation closes. Lower means faster turnover (a café); 1.5 is recommended. |
Behaviour switches
| Switch | Set it for |
|---|---|
| Face match required | On for banking: if the ID chip checks out but the face does not match, the person is refused. |
| Greet when a person appears | Welcomes new visitors automatically and greets recognised visitors by name. Needs camera permission and a visible face. |
| Register by name (ID not required) | On for a café or a shop — an unrecognised visitor is asked their name and registered by voice. Off where identity must be proven: people are identified only through the ID card. |
| Send kiosk camera to the cloud | Only when the FaceID server has no camera of its own. Leave it off if FaceID and the camera run locally. |
The greeting scripts
Seven fields under Greeting scripts, each spoken by the kiosk exactly as written (they do not go to the language model), with {name} replaced by the visitor's name. Clear a field and save it to skip that greeting.
| Field | Example line |
|---|---|
| Recognized person (face) | "Welcome back, {name}. Good to see you again." |
| Unrecognized person (face) | "Hello there. I don't think we've met — what's your name?" |
| Request ID (registration off) | "Could you hold your ID card to the reader for me?" |
| Verified returning person | "You're already verified, {name}. How can I help?" |
| Identity verified (NFC) | "Thank you, {name}. You're verified." |
| Face didn't match (NFC) | "I'm sorry, the photo on the card doesn't match. Please see a colleague at the desk." |
| No live face (NFC) | "I can't see you clearly — could you look at the camera?" |
Phone — give it a phone number
Attach a real phone number and the character answers calls in its own voice, with all of its knowledge, memory and tools intact. On Channels › Phone, the Inbound calls card holds Enabled, Phone number (E.164), Greeting and the carrier's Webhook URL; the Outbound call card dials a number to test it. Full setup is on the Phone calls page.
WhatsApp — answer on WhatsApp Business
On Channels › WhatsApp you set how this character behaves on WhatsApp: Enabled, Greeting, Reaching a person, Voice replies and Read receipts and typing. The Number is read-only here — numbers are connected and assigned in My workspace › Settings › WhatsApp Business. Open the call centre goes to the live WhatsApp inbox. A character that uses a realtime voice model (Gemini Live or OpenAI Realtime) cannot answer on WhatsApp. Full guide: WhatsApp Business.