Microsoft says AI voice technology still needs improvement
1 week ago / Read about 0 minute
Author:小编   

In April 2026, Mustafa Suleyman, head of Microsoft AI, stated that artificial intelligence still has a long way to go before achieving a truly natural experience with voice commands, and that models and agents require extensive training to accurately understand human intentions. He made these remarks while discussing Microsoft's new voice transcription model, MAI-Transcribe-1. The model supports 25 languages, with an average word error rate of just 3.9% in the FLEURS benchmark test, and its batch transcription speed is 2.5 times that of the existing Microsoft Azure Fast service. However, the initial version does not support features such as real-time transcription or speaker separation.