MOSI AI is advancing human-computer interaction through AI research and products. We build real-time voice and multimodal systems for developers, researchers, and businesses.
Our work spans audio understanding, speech generation, and deployment-first multimodal infrastructure. Together with the OpenMOSS team, we have open-sourced MOSS-Audio, MOVA and our MOSS-TTS family, including MOSS-TTS-Nano, an ultra-lightweight multilingual model designed for real-time streaming, CPU inference, and practical deployment.
Our MOSI Studio empowers creators and brands to create and edit real-time voice, video, and multimodal content with context-aware AI.
Our MOSI API enables businesses and developers to empower their customer interactions and experiences with our leading models.
These systems are powered by our research in low-latency speech generation, audio tokenization, and multimodal modeling. We focus on making advanced AI more responsive, more deployable, and more useful in real-world products and devices.
