Documentation
One photo becomes an avatar; your documents become its answers; one line of code puts it on your site. Everything the dashboard does, the API does too.
How the pieces fit
- Avatar: a face built from one photo (or a Studio image). It is rendered live on our GPUs and streamed to the visitor’s browser as video frames and audio.
- Agent: an avatar plus a voice, a charter (purpose, tone, greeting, boundaries), knowledge files, and limits. Agents are drafted, then published; visitors always get the published version.
- Embed: a publishable id for a published agent, locked to the domains you list. The widget, the hosted page and the iframe all use it. Keys never go in a browser.
- Session: one conversation. The browser gets a short-lived WebSocket URL and token, dials the realtime relay, and streams microphone audio or typed text up and frames, audio and transcripts down.
- Minutes: a live avatar on screen draws one minute per minute; a Premium avatar draws two; Studio operations draw one to three per image. Extra minutes past your plan run under a spending cap you set.
Quickstart: five steps
- Sign in at app.avaclone.ai. A new workspace starts on the free trial with 15 minutes.
- Build an avatar under Avatars: upload a front-facing photo, or pick a Studio image. A build takes about a minute on a warm engine, a few minutes on a cold one.
- Create an agent: choose the avatar and a voice, write the charter, drop in PDFs or text as knowledge. Test it in the Test tab, then Publish.
- Add your domain in the Embed tab (the domain the widget will run on) and copy the snippet.
- Paste the snippet before
</body>:
<script src="https://cdn.avaclone.ai/v1/widget.js" data-agent="pub_…" async></script>That gives a launcher in the corner. For an avatar inside your own layout, give it a target element:
<div id="avatar" style="width:380px;height:620px"></div>
<script src="https://cdn.avaclone.ai/v1/widget.js" data-agent="pub_…" data-mode="inline" data-target="#avatar" async></script>No site to edit? Every published agent also has a hosted page at app.avaclone.ai/a/<publishable id> that you can link to or put in an iframe. See the widget page for every option.
Going further
- Browser SDK: mount the avatar in your own UI, drive it with your own microphone capture or text, listen for transcripts and state.
- REST API: create avatars, agents, knowledge, embeds and sessions from your server with a secret key. Generate and restyle Studio images.
- Changelog: what changed and when.
Limits worth knowing
- Live conversations at once, avatars and embed domains are per plan; see pricing. Starting a conversation past the cap returns
429 cap.live; out of minutes returns429 minutes.exhausted. - An agent’s own limits: turns per hour and per day for one visitor, and a quiet cut-off (by default two minutes of silence ends the session so idle tabs do not burn minutes).
- Session tokens expire in 10 minutes if unused; a session may run for about an hour.
- Every avatar is labelled as AI in the widget, says so when asked, and never claims to be the person it looks like. An avatar of a real person needs that person’s signed release first.