Kaltura Gives AI Avatars an Emotional Layer
The Israeli video technology company launches avatars that analyze tone, intent, and conversational context in real time, aiming to make AI interactions more natural across sales, customer service, and training
Photo: Kaltura
The next challenge for conversational AI is no longer just teaching machines what to say — but how to say it.
Kaltura is introducing a new capability for its Agentic Avatars that allows AI-powered digital characters to interpret a user’s tone, conversational intent, and emotional cues during an interaction, and respond with synchronized facial expressions, vocal changes, and adaptive behavior.
The move comes as companies increasingly deploy AI agents and avatars across customer service, marketing, sales, employee training, and education. However, as adoption grows, organizations are discovering that simply providing automated answers is not enough. The quality of the interaction itself can determine whether users remain engaged or quickly abandon the experience.
Unlike traditional avatars that rely on fixed expressions or scripted responses, Kaltura’s avatars dynamically adjust their behavior throughout a conversation. They can display empathy, concern, enthusiasm, surprise, or skepticism depending on the context and the way a user communicates.
The technology is designed for a range of enterprise applications. In customer support, an avatar that detects frustration in a user’s voice can respond with a more reassuring approach. In sales, it can adapt its style based on a buyer’s level of interest or hesitation. In training and learning environments, avatars can provide encouragement when users struggle or adjust their delivery when confusion is detected.
Kaltura is also using the technology in roleplay simulations, where employees can practice difficult conversations with AI characters acting as skeptical customers, frustrated users, or demanding executives. Unlike scripted simulations, the avatars respond dynamically as the conversation develops.
The underlying technology is based on Kaltura’s proprietary diffusion rendering architecture, which uses predefined emotional reference points and blends between them in real time to create smoother facial expressions and reactions.
“Human conversations rely on much more than words,” said Dr. Alan Bekker, CTO at Kaltura. “Tone, facial expression, and emotional context all influence how people interpret an interaction. By enabling avatars to respond to those signals in real time, we are making conversational AI feel significantly more natural.”
The launch follows Kaltura’s broader expansion into agentic AI experiences, including its acquisitions of conversational avatar company eSelf.ai and AI-powered content intelligence platform PathFactory.