Xiaomi Unveils Advanced AI Voice Model for Cars and Smart Homes
Xiaomi has launched its new AI voice model called MiDashengLM-7B, designed to enhance smart home and automotive experiences by enabling voice control with high responsiveness and offline capability. The model supports more than 30 smart features, including controlling home appliances like air conditioners, TVs, and lights, as well as car functions such as music, air conditioning, and window operations. It integrates advanced voice recognition and natural language understanding, allowing it to accurately interpret commands even from across the room or in noisy environments.
Next-Gen AI with Wide-Ranging Capabilities and Open Source Availability
The MiDashengLM-7B model is distinctive because it not only processes speech but also comprehends ambient sounds, background music, and environmental noises. This capability comes from combining Xiaomi’s Dasheng audio encoder with Alibaba’s Qwen2.5-Omni-7B decoder, enabling gesture controls through claps or snaps and enhancing security with 24/7 audio monitoring and alerts for unusual noises. It also features real-time language pronunciation feedback for drivers practicing new languages. Additionally, it includes a unique underwater voice wake-up function for wet environments. Another highlight is its energy-efficient design, which reduces response delays and handles a significantly higher number of simultaneous requests without extra memory use, enabling offline functionality.
Midashengglm-7b of Xiaomi is published as an open-source model under the Apache 2.0 license, encouraging developers and companies to integrate this AI in their applications for research and commercial use. This strategic initiative aims to establish a robust developer ecosystem centered on Xiaomi's artificial intelligence technologies, fostering innovation in the intelligent home and automotive industries. The AI model is currently implemented in the Xiaomi product line in China and is designed to make users' lives more intelligent, more comfortable, and safer through voice-centered control.