Avatars Gain 'Emotional Intelligence': Recognizing Tone Beyond Words
Israeli company Kaltura has launched AI-powered avatars capable of analyzing speech tone, intent, and emotional context in real time. The technology aims to make corporate training simulations more natural and effective.
More and more companies are moving their customer service to AI-based bots, but conversations often hit a dead end when the bot fails to understand the situation and responds inappropriately. Now, the Israeli company Kaltura is launching AI-powered avatars with emotional intelligence. These avatars can analyze speech tone, intent, and emotional context in real time, adjusting their facial expressions, vocal tone, and response style accordingly.
The system is available in 40 languages, including Hebrew, and is primarily intended for organizations using simulations for training, professional development, and sales. For instance, a customer saying 'thank you very much' might be genuinely grateful or, conversely, sarcastic. While a human operator understands the intent naturally, it remains a complex challenge for AI. An incorrect response can turn a routine conversation into an unnecessary confrontation.
'In the past, both situations could have received a similar response, but today the system tries to distinguish between them and respond accordingly,' says Dr. Alan Becker, Kaltura's CTO.
Kaltura's avatars analyze not just the words spoken, but also intonation, pace, hesitations, emphasis, and conversational context. The product is designed for companies training support representatives, managers, and HR teams, as well as organizations in health and education. 'Instead of practicing against a fixed script, employees practice against an avatar that reacts dynamically, just like a real person,' Becker explains.
The system utilizes Kaltura's proprietary AI technology, allowing avatars to change facial expressions in real time. The avatar can express joy, empathy, concern, surprise, or skepticism, controlling the intensity of the expression as the conversation evolves. According to Becker, the development was driven by customer feedback, as users want to feel they are talking to someone who understands the situation, not just the words.
'The goal is not to 'read minds,' but to estimate the intent in a probabilistic way and respond in a way that is more suitable for the situation,' he adds. Becker does not expect the technology to replace humans: 'AI will handle a large part of routine inquiries, but humans will continue to handle complex, sensitive cases that require judgment and human empathy.'
