Although the quality of TTS has improved considerably over the past few years, many commercial enterprises prefer to use pre-recorded speech in order to guarantee high-quality output. Professional artists, often referred to as voice talent, are employed to record the system's prompts.
The downside of pre-recorded prompts is that they cannot be used where the text to be output is unpredictable—as in apps for reading e-mail, text messages, or news, or in applications where new names are being continually added to the customer list. Even where the text can be predicted but involves a large number of combinations—as in flight announcements at airports—the different elements of the output have to be concatenated from pre-recorded segments but in many cases the result is jerky and unnatural. Another situation is where output in other languages might be made available. It would be possible to employ voice talent to record the output in the various languages...