AvatarTalk builds talking avatars you can hold a real conversation with. The same platform answers customers from a tablet at your entrance or a widget on your website, generates avatar video from text through an API, and plays a realistic counterpart when you rehearse a difficult conversation at work.
What we run
Speech recognition, language models and avatar video rendering are run as one hosted service, so a tablet at the entrance needs nothing more than a browser and an internet connection. The same stack can run on an embedded appliance when the venue needs the avatar to keep working without depending on a network link.
One account and one credit balance cover every surface. One credit is one second of avatar video, charged the same whether it was spoken at a kiosk, answered in a widget or returned by the API. Failed generations are not charged.
How we handle what you put in
We do not store audio or video. Microphone audio is processed to produce text while a session is live, and avatar video is streamed to the browser in front of the visitor. For customer assistants we keep only the timing and outcome of a conversation, so billing never depends on what was said. The full detail is in the privacy policy.
Contact
Write to support@avatartalk.ai. One address reaches us for everything: questions before you start, help with a deployment, billing, privacy and data requests, or security reports.
Tell us what the assistant should handle and where it will stand, and we will suggest the simplest setup rather than the largest one.
Who owns the service
AvatarTalk is owned and operated by AvatarTalk, Stanisława Konarskiego 18C, Gliwice, Poland.
Want to see it answering your own questions?
Build an assistant