An avatar you can talk to.

Athira avatars are drawn live in the browser, speak in real time and respond to the conversation as it happens. Ask Maren something and see for yourself.

Talk to avatarExplore the Athira API
Maren, an Athira avatar, her portrait breaking up into dots on one side

Talk to Maren.

Pick a question to start, or type your own. She answers the way she would on a company website.

Start with a question

    Four systems, one reply.

    Everything Maren does comes from four systems working together. Here they are in a single answer.

    A visitor on a company website

    "We're looking at this for our support team. Can someone show us how it works?"

    Maren

    Happy to. For a support team, the practice partner for customer calls is a good place to start. Shall I book a demo with the team?

    1. 01

      Context

      What she knows.

      She answers from the knowledge and instructions the company approved, so she knows which part of the product fits a support team.

    2. 02

      Behavior

      How she responds.

      She reads this as a request to see something, not a question to explain at length, so she keeps the answer short and warm.

    3. 03

      Voice and visual

      How she says it.

      Speech streams out while the reply is still being made. Each mouth shape is timed to the sound it belongs to, and she blinks, breathes and nods between words.

    4. 04

      Actions

      What she does next.

      She offers the next step, and books the demo when the visitor says yes.

    Receptionist and guide.

    The first face people meet on your website, and the one who shows them around. The same avatar also teaches and plays a practice partner.

    A made-up energy company homepage with Maren in the corner, greeting the visitor
    A made-up finance app setup screen with Maren explaining the approval limit
    A made-up college visit page with Maren offering to book a campus tour, and three time slots

    Example sites with sample content. Northvale, Ledgerly and Easton College are made up.

    Built for a close-up.

    Her face, eyes, mouth, hair and clothes are separate layers, drawn live in the browser on every frame.

    Brows and lashesBlend shapes on an ARKit-52 rig move the mouth, eyes and brows as she talks.
    EyesTheir own parts, so she can look at the person she is talking to.
    Teeth and tongueShaped for speech, with lip sync timed to every sound.
    HairA separate layer, so a look can change without touching the face.
    Face and bodyA dense surface, built for close-up speech.
    ClothingA separate layer too, like the hair.
    Loading the model…
    • Brows and lashes Blend shapes on an ARKit-52 rig move the mouth, eyes and brows as she talks.
    • Eyes Their own parts, so she can look at the person she is talking to.
    • Teeth and tongue Shaped for speech, with lip sync timed to every sound.
    • Hair A separate layer, so a look can change without touching the face.
    • Face and body A dense surface, built for close-up speech.
    • Clothing A separate layer too, like the hair.

    Fast enough to feel like a conversation.

    Speech is transcribed while the person is still talking, the reply is decided against the task at hand, and audio streams out while it is still being made.

    Until a reply beginsResponses typically begin within a second.
    Face shapes on the rigARKit-52 blendshapes drive speech and expression.

    English is supported today. More languages are planned.

    Where to go from here.

    Talk to MarenAsk her another question, or try one you would expect from your own visitors.Build with the APIPut the avatar on your website or in your product. Plans and docs are on the Athira API site.Request a custom avatarMaren, Aris and Sena are ready now. Your own face, voice and name are made on request.