Request a demo
Core Avatar

The real-time avatar primitive for whatever you are building.

Core Avatar gives your application a lifelike face, voice, and conversational presence without prescribing the job it should do. Bring your own intelligence, knowledge, and workflow, then deploy through Athira's API and embeds.

Read the documentationContact Sales
Core Avatar render
Five open sockets

Athira carries the conversation.
Your product decides it.

Intelligence

Your models and routing decide what the avatar says - it carries the reply, never chooses it.

Workflow

Your conversation logic defines what happens turn by turn, and where each path leads.

"Walk me through how you'd handle a customer whose warranty expired."
Your intelligence
Retrieving warranty policyyour sources
Routing to your modelyour choice
Speech streaming outAthira
Ticket written backyour systems

Identity

Appearance, name, and voice come from you - the primitive ships without a persona.

Knowledge

Your retrieval and sources ground every answer. Nothing is assumed about your business.

Actions return to your systems
No prescribed job or persona
Provider-agnostic pipeline
API and embed surface
See it in actionReview security
By design

It ships empty. That is the point.

Core Avatar arrives with no job, company identity, business knowledge, or visitor workflow. Your product defines those layers - Athira provides the real-time human interface.

Filled by AthiraFace, voice, and real-time presence
Empty
No assigned job
Empty
No company identity
Empty
No business knowledge
Empty
No visitor workflow
Ownership boundary

It is not a list of settings. It is the line between two responsibilities.

What Athira supplies
The interface

Everything a person sees and hears, rendered live.

In-browser avatar deliveryRendered live from current state, no plugins
Real-time two-way audioStreaming in and out on the same turn
Speech and lip syncPhoneme-timed visemes aligned to the audio
Expression and micro-motionTone, blinking, breathing, and idle motion
API and embed surfaceSessions, personas, events, and placement
What you control
The decisions

Everything that decides what is actually said and done.

Intelligence and routingWhich model answers, and how it is chosen
Knowledge and retrievalThe sources every answer is grounded in
Identity and voiceWho the avatar is and how it sounds
Conversation logicWhat happens turn by turn, and where it leads
Downstream actionsRecords, bookings, tickets, and next steps

Athira never chooses what is said, only how it is delivered.

Rendered live, frame by frame.

The avatar is drawn live, frame by frame, from whatever is happening in the conversation - so it can pause, react, and pick up mid-sentence like a person would. Technical detail is in the documentation.

Drawn fresh, every frameEach frame is generated from what is happening right now - not a clip picked from a library.
Not a video, not a loopNothing is pre-recorded, so the avatar can react to anything mid-sentence.
Looks like a person, not a renderSkin, eyes, and light are handled so the face reads as human rather than as a graphic.
Runs in the browser they already haveNo download, no plugin, and nothing to install for the person on the other side.

Speech and behavior.

Every turn moves through the same four stages. You choose what sits in each one.

01Athira
Speech to text

Streaming transcription as the person speaks.

02Yours
Reasoning

Your intelligence produces the reply.

03Athira
Text to speech

Audio streams out while it is still being made.

04Athira
Face and body

Visemes and motion resolve against the audio.

Mouth
·Phoneme-timed visemes aligned to the outgoing audio
·Coarticulation, so sounds blend the way real speech does
·Emphasis that follows the stress in the sentence
Face and body
·Expression that carries the tone of the reply
·Blinking and breathing between turns
·Subtle sway, nods, and head tilts while listening

Integration pattern.

Your application stays in charge. Athira renders the conversation, and every action comes back to you.

Yours
Your application

Your product, surface, and session.

Yours
Your intelligence

Models and retrieval you route.

Athira
Athira Core Avatar

Face, voice, and real-time presence.

Yours
The conversation

What the person sees and hears.

Your application

Your product, surface, and session.

Your intelligence

Models and retrieval you route.

Athira Core Avatar

Face, voice, and real-time presence.

The conversation

What the person sees and hears.

Actions return to your application - bookings, records, tickets, next steps.

Frequently
Asked
Questions

Give your intelligence a human interface.

View documentationTalk to Sales