Real-time Audio-Visual Interaction, Millisecond Response: "Riri Xian" Big Model Interactive Platform Joins Xiaomi AI Glasses

China Hi-Tech News August 6, Shanghai Tech announced that its "Riri Xian" big model interactive platform has successfully connected with Xiaomi AI glasses, enabling users to achieve seamless integration of looking, speaking, remembering, and thinking in real-life scenarios, with the ability to perform bilateral real-time audio-visual interaction.

According to reports, traditional intelligent devices' interaction experiences are often limited by response delays, context breaks, and one-way input. The highlight of Xiaomi AI glasses combined with Shanghai Tech's "Riri Xian" is its bilateral real-time audio-visual interaction capabilities, which go beyond simple voice commands or semi-bilateral concatenation, but rather build a natural, smooth, and uninterrupted dialogue loop, such as real-time recognition during street exhibitions, becoming a Q&A small encyclopedia; and when traveling abroad, becoming a translation assistant, showcasing powerful capabilities in various scenarios:

Millisecond response, thinking is communication: relying on the strong model reasoning ability and underlying optimization of "Riri Xian", interaction delays are compressed to milliseconds, users' speech has not fallen, understanding has begun, and responses are almost real-time generated.

Contextual continuity without interruption: "Riri Xian" big model can deeply understand the context above and below, precisely track the dialogue thread, support interrupting, correcting, and deepening questioning, making conversations like interacting with a true assistant:

Complex environment feedback super stability: even in noisy exhibition halls or bustling streets, the noise performance upgrade can ensure that commands are accurately captured and understood without error.

Deep analysis, memory support: combining audio-visual memory and retrieval technology, the system can immediately associate historical communication details (such as recalling a customer's plan), providing extremely targeted information support:

According to previous WAIC 2025 big model forum information released by Shanghai Tech, after the "Riri Xian V6.5" big model update, its interaction performance has been significantly improved, surpassing Gemini 2.5 Flash and GPT-4o in terms of multimodal interaction capabilities.