Siri’s New Voice: Apple’s AI-Powered Update
Siri’s new voice, now available in iOS 11, marks a significant step forward in the evolution of virtual assistants. The traditional method of creating synthesized speech, based on piecing together pre-recorded sounds, has been replaced by a more advanced machine learning approach.
Apple spent over 20 hours recording a new voice talent, using a vast library of scripted speech from various sources, including books, jokes, and question-and-answer sessions. This speech data was then broken down into minuscule sound fragments called “half-phones,” which are even smaller than the traditional “phones” used in speech synthesis.
These half-phone fragments were then fed into a machine learning model capable of determining the most appropriate sound for each situation, taking into account elements like sentence start and end, tone, and speed. This approach results in a more natural-sounding voice that replicates the nuances and subtleties of human speech.
The new Siri voice not only sounds less robotic but also conveys a greater sense of energy and personality. Speech experts have even identified a distinct Californian accent in Siri’s new voice, highlighting the level of detail achieved through the machine learning process.
Siri’s new voice, now available in iOS 11, marks a significant step forward in the evolution of virtual assistants. The traditional method of creating synthesized speech, based on piecing together pre-recorded sounds, has been replaced by a more advanced machine learning approach.
Apple spent over 20 hours recording a new voice talent, using a vast library of scripted speech from various sources, including books, jokes, and question-and-answer sessions. This speech data was then broken down into minuscule sound fragments called “half-phones,” which are even smaller than the traditional “phones” used in speech synthesis.
These half-phone fragments were then fed into a machine learning model capable of determining the most appropriate sound for each situation, taking into account elements like sentence start and end, tone, and speed. This approach results in a more natural-sounding voice that replicates the nuances and subtleties of human speech.
The new Siri voice not only sounds less robotic but also conveys a greater sense of energy and personality. Speech experts have even identified a distinct Californian accent in Siri’s new voice, highlighting the level of detail achieved through the machine learning process.
