Interactive AI avatar · from $0.02/min
An interactive avatar is an AI character on screen, often a photorealistic person, that holds a live spoken conversation: it listens through the user's microphone, answers in a natural voice, and its lips and expressions move with every word, in real time. Talking Avatar gives you one with no code, or adds one to the voice agent you already run.
MayaYour order shipped this morning. Want me to text you the tracking link?
Under the hood, an interactive avatar is a voice agent with a face: speech recognition hears the user, an AI model decides the reply, a voice speaks it, and an avatar model turns that voice into a moving face. The names below are used for the same idea.
| Question | Interactive avatar | Avatar video |
|---|---|---|
| What the user does | Talks to it | Watches it |
| When the video is made | Live, during the conversation | In advance, from a script |
| What it says | Decided turn by turn by an AI model | Written before it's rendered |
| Typical uses | Support, sales, interviews, tutoring | Marketing, training and explainer videos |
Talking Avatar is for live conversations. It doesn't render videos from a script.
No code · $0.06/min, all-in
The whole interactive avatar, ready to share: you write the prompt, we run the AI model, speech recognition, voice and avatar.
For developers · $0.02/min
Already run a voice agent? Keep your own AI model and voice, and give it a photorealistic face.
Both are billed per second from prepaid credits, with no subscription. New accounts get $1 of free credits, with no card needed. See full pricing
In their web browser, with their microphone, on desktop or mobile. There's nothing to install.
With Voice Agents, our realtime AI model follows your prompt and can call your tools. With the API, your own agent does it.
A natural voice speaks the reply, streamed as audio while it's generated, so there's no wait for the whole answer.
Our servers render the lips, expressions and subtle head motion from that audio, and stream video and voice together over WebRTC.
The user can interrupt at any time: the avatar stops mid-sentence and listens, like a person would. Video streams at 832×468, 25 fps, and the lips, expressions and voice arrive together, so they stay in sync.
The avatar starts speaking each reply less than 250 ms after the voice reaches it, so the face keeps up with the conversation.
<250ms
Avatar response time
Need a free avatar that runs entirely in the browser instead? Talking Avatar Open Source animates your own 3D .glb model from any audio, with no server or API key. It isn't a photorealistic person. See the open-source SDK
An interactive avatar is an AI character on screen, often a photorealistic person, that holds a live spoken conversation: it listens through the user's microphone, answers in a natural voice, and its lips and expressions move with every word, in real time. Unlike an avatar video, nothing is scripted or recorded in advance: an AI model decides each reply, and the face is rendered live.
Usually. "Digital human" tends to mean an interactive avatar that looks like a real person. Talking Avatar Realistic Avatars are digital humans in that sense: photorealistic people who talk in real time.
With Talking Avatar, $0.06 a minute for a complete no-code agent (avatar, AI model, speech recognition and voice), or $0.02 a minute for the avatar alone through the API. Both are billed per second, with no subscription. A custom avatar is $199 one time.
Yes. Talk to one live for two minutes on the demo page, with no signup. New accounts get $1 of free credits, with no card needed: 50 minutes through the API, or about 16 minutes with a voice agent.
Not with Voice Agents: you describe the agent in the dashboard, then share its link or paste one line on your site. The Realistic Avatars API needs a developer, who streams your agent's audio to the avatar from your server.
Yes, through the Realistic Avatars API. The avatar follows audio from any voice agent or TTS voice, and can join your LiveKit room as a participant.
A voice agent understands callers in their own language, and replies in it when you pick a Multilingual voice. Through the API, the avatar follows your agent's audio in any language.
No. People talk to it from a web browser with a microphone, on desktop or mobile. Phone calls and phone numbers aren't supported.
A free two-minute call in your browser, with no signup. Then build your own with $1 of free credits.
Talking Avatar