Sharing is caring!

Kyutai, a non-profit AI research lab, has unveiled Moshi, the first voice-enabled AI model accessible to the public. Developed in just six months by a team of eight, Moshi allows for smooth, natural, and expressive interactions with AI. During its presentation in Paris, attendees including researchers, developers, and journalists interacted with Moshi, demonstrating its potential as a digital coach or companion, and its creative abilities in role-playing scenarios. Moshi’s capabilities include exceptional text-to-speech functions, enhancing emotional and interactive communication between multiple voices. The AI model can be tested online via Kyutai's website, marking a significant milestone for generative voice AI technology. Kyutai aims to contribute to open AI research and development by making Moshi’s code and model weights freely available. This initiative supports researchers and developers in advancing voice-based products and services. Compact and capable of running on unconnected devices, Moshi represents a breakthrough in AI safety and accessibility. Kyutai, founded in November 2023 by the iliad Group, CMA CGM, and Schmidt Sciences, focuses on developing high-capability general-purpose models. The lab emphasizes multimodality, utilizing diverse content types for learning and inference, with all developed models and related software shared openly to foster innovation in the AI community.

Visit
Find us on AI Scores

Sharing is caring!