Tavus positions itself as "The Human Computing Company" and has been committed since 2020 to building AI virtual humans—PALs (Personified Application Layers)—that can see, hear, respond in real time, and communicate face to face. A PAL is not just a face and a voice, but an AI character with full interaction capabilities: it can express emotions and move naturally; it has a persona, personality, and rules assigned by the user, and can remember every conversation; it speaks with pacing, warmth, and intonation similar to a real person; out of the box, it can present slides, instantly create visual content, share its screen, and join Zoom or Google Meet calls; it can also read the user's expressions, attention, and surrounding environment, understand the full context, and distinguish tone, hesitation, pauses, and complete thoughts, thereby judging when to interject and when to leave space.
In terms of how it is used, Tavus offers three paths. Developers, founders, and teams can integrate real-time, face-to-face AI conversations into products through the API, building advanced and deeply customized PAL experiences. Non-technical users can use PAL Maker, with no code or engineering background required, to create and publish a PAL in minutes by describing the task, assigning a persona, choosing a face, and configuring a knowledge base. Enterprise customers can choose a fully managed solution, with the Tavus team designing, building, tuning, and scaling from concept to production to create custom PALs that meet business needs.
The platform emphasizes "frontier models powering seamless conversation." Its preview Griffin is billed as the first human interaction model: a video-to-video model that can simultaneously see, hear, speak, and move like a person. It keeps listening while speaking, reacts to what the user shows it, and gestures while speaking, enabling a machine for the first time to take on one side of a face-to-face exchange. Note, however, that Griffin-Lite is currently not available to customers, and Tavus provides a request channel for selected trusted testers; official release timing, commercial pricing, and customer API access have not yet been determined by this preview.
Applicable scenarios are broad, including sales development, coaching and training, continuing medical education, math tutoring, interviews, concierge services, and companionship. Public cases show that Salesforce built Piper, a real-time AI sales development representative, on Tavus, handling Salesforce's main touchpoints and conversing with tens of thousands of visitors each day; Orum built Aurora, an AI sales role-play coach, for representatives to practice realistically at any time; Med Learning Group built Ruby, billed as the world's first cSCC AI clinical advisor, for continuing medical education, with answers drawn from vetted sources. Tavus says these deployments have produced results such as increased meeting scheduling, higher conversation engagement, shorter representative ramp-up time, improved conversion rates, and increased coaching revenue.
