The Longgang-based AI startup Amphion has achieved breakthroughs in speaker identification and AI-powered voice anti-spoofing, propelling context-aware voice intelligence from the laboratory to commercialization and laying the foundation for human-machine interaction in the era of physical AI.

Amphion focuses on the research and development of foundational models for context-aware intelligent voice, aiming to build the next-generation voice interaction infrastructure for the era of physical AI.

Professor Wu Zhizheng, a scholar at the Chinese University of Hong Kong, Shenzhen, founded the company. The team possesses two decades of accumulated experience in AI voice R&D and industrialization.Core members have achieved multiple results that rank among the industry's top international benchmarks.

With just three to five seconds of enrollment audio, the AmphionASR personalized speech recognition system locks onto the target speaker, transcribing only the designated voice while filtering out all other speakers in a crowded, multi-talker environment.

Amphion's Software Development Kit has now been deployed on several hundred thousand devices, with expansion underway toward the ten-million-unit scale. Its voice anti-spoofing technology has also transitioned to full commercial deployment.

The AI ecosystem community Molidu OPC — that driving force behind Amphion — is housed within the Longgang Venture Capital Building.Spanning 10,000 square meters, the space is a joint establishment by Longgang district and the China Academy of Science and Technology Development Institute.