Abstract: Due to the needs of developing Computer Aided Instruction (CAI) we built a Cantonese (Guang Zhou Language) text-to-speech system based on IBM microcomputer. We studied the sound production rules of Cantoness and its acoustio parameters. The system used Formant Synthesizer and selected 578 syllables and six tones as the speech elements. The acoustio parameters of each speech element were extracted from the runing speech. We built an acoustic parameter database. In order to achieve the transform from text-to-speech we designed a phonetic management program. The system has output spoken Cantonese of both Chines words and some kind of symbols (English, Japanese, Russian or Science Symbols), and freely control the word displayed and sounded by the computer.