Towards Interactive Intelligence for Digital Humans

Abstract

We introduce Interactive Intelligence, a digital-human paradigm that enables personality-aligned expression, adaptive interaction, and self-evolution. To realize this, we present Mio (Multimodal Interactive Omni-Avatar), an end-to-end framework with five modules: Thinker, Talker, Face Animator, Body Animator, and Renderer. This unified architecture combines cognitive reasoning with real-time multimodal embodiment for fluent and consistent interaction. We also establish a dedicated benchmark to evaluate interactive intelligence capabilities. Extensive experiments show that Mio outperforms strong baselines across all evaluated dimensions.