Tavus, a company specializing in Conversational Video Interfaces (CVI), has announced the launch of three new AI models: Phoenix-3, Raven-0, and Sparrow-0. These models are designed to revolutionize human-AI interaction by creating more authentic, face-to-face experiences.
Introducing our new family of state-of-the-art AI models: Phoenix-3, Raven-0, and Sparrow-0.
— Tavus (@heytavus) March 6, 2025
Together they bring Conversational Video Interfaces (CVI) to the next level, and power Charlie, our new demo persona 👋 pic.twitter.com/vaqX9InvwX
Phoenix-3 is an advanced full-face animation model that captures facial expressions, micro-expressions, and lip synchronization with remarkable accuracy. It enables continuous, context-aware emotional responses while preserving identity fidelity.
Raven-0 provides AI with visual perception capabilities, allowing it to interpret and respond to visual cues and ambient context in real-time. This model can read text, recognize gestures, and detect emotions, making interactions more natural and intuitive.
Sparrow-0 focuses on improving the timing and flow of conversations. It adapts to natural speech patterns, understanding the rhythm and context of dialogue. This model outperforms current state-of-the-art systems in Turn Accuracy, achieving a Mean Absolute Error of 0.3989 compared to the previous best of 1.7467.
These models work together to form an intelligent chain-of-thought system within Tavus's Conversational Video operating system. The company has demonstrated the capabilities of these models through an AI agent named Charlie, which can engage in lifelike dialogue, search the internet, analyze screens, and generate images seamlessly.
Tavus is making these models publicly available via their APIs on March 6th, 2025. The company envisions widespread adoption of this technology, with digital humans becoming as common as smartphones. Several leading brands and startups are already integrating this technology into their workflows.
