Pick the face and how it ships. The preview is live; every change updates the copy-paste snippets and this page's URL (share it — the configuration rides the query string).
LAB faces (samples, hybrids, photo-real) need the embed served same-origin with the Alive site, or CORS on the host. Cast faces work anywhere.
Pairing with a voice? Match the face: ♂ faces suit deeper voices (cedar-class), ♀ faces suit marin-class — mascots, creatures and robots pair with anything. The ★ hybrids are the signature look.
Captions + accessibility follow this (incl. right-to-left). The VOICE speaks whatever language your model is told to speak — modern voice models are natively multilingual; put the language in your agent's instructions (NetShow-hosted agents: one line in the console's AGENT BRAIN).
Corner + takeover need zero markup — the tag alone mounts them.
Exposes embed.express() (whitelisted, rate-limited). Wire your model's express_self tool calls to it; without a session driving it, nothing changes.
LIVE PREVIEW · SCRIPTED, CAPTIONS ONLY, $0 · THE RENDERER RUNS IN YOUR BROWSER
A purchased install ships with a config id instead of hardcoded styling: their tag says data-config="<id>", the styling lives in a file WE host (embed/cfg/<id>.js). Update that file and every install of theirs restyles instantly — no customer code change, no support ticket. Hosted styling wins over their tag; placement (inline/corner/size) stays theirs.
embed/cfg/<id>.js)