Digital Avatar software suite

A photorealistic talking avatar that holds a voice dialogue in real time: it hears the question, answers and moves its face in sync.

What it can do

An answer without a pause

the first sentence goes to speech synthesis and animation without waiting for the generation to finish
A face from a photograph

the avatar is built from a single portrait shot or a short video
Interchangeable speech modules

recognition, synthesis and the language model are chosen separately: cloud or fully local
Behaviours

greeting, waiting, listening, thinking, speaking; configured for each avatar
Eight emotions

neutral, joy, sadness, anger, fear, surprise, disgust, contempt; switched on the fly
Function calls

the language model reaches into your system: open a document, run a query, fetch information
Web search

an optional mode in which the answer is supplemented with fresh data from the internet

Greeting: the avatar waves when it connects

Idle: a neutral pose between lines

Speech: lips and expression stay in sync with the sound

How to show it

In a browser

audio and video go over WebRTC, with nothing to install
Kiosk application

a separate client for Windows and Linux, full screen without a browser
Service channel

the statuses listening and thinking, and commands for the kiosk interface

Who it suits

Museums and exhibitions

information kiosks in public spaces
Corporate stands and reception desks

greeting guests, wayfinding, answers to frequent questions
Educational platforms

virtual teachers and mentors
Voice support and wayfinding

Platform

A Linux server with an NVIDIA graphics card

from 8 GB of video memory
Deployment in containers

Docker and Docker Compose
Nothing sent to the cloud

with local speech modules the data never leaves the perimeter of the client

Projects

Request a quote

Discuss a project