NEW YORK (AP) _ AT&T Corp. introduced a speech simulation system Tuesday that recreates the sounds of human voices from text.

Other technologies can also produce spoken words from written text, but AT&T says its new system produces more natural-sounding voices that can also mimic the voice of any particular person.

AT&T hopes to sell the ``text-to-speech'' technology to companies that operate call centers and telephone services such as travel or reservation systems.

The new product from AT&T Labs, Natural Voices, consists of two parts: an ``engine'' that translates text into natural-sounding speech, and a library of voices that can be used, including ones that could mimic the voices of well-known people.

IBM and Lernout & Hauspie Speech Products also make products that translate text into speech.