A text-to-speech system was developed which reads aloud arbitrary German texts. It converts an orthographic input text into synthetic speech in 3 main steps: At first the words of the text are preprocessed and abbreviations, special characters, and digits are replaced by their full orthographic correlates. In a second step the preprocessed words are phonetically transcribed by means of about 1,370 letter-to-phone rules. The rate of incorrect transcriptions is near 2% of the running words of a text. Finally the phone symbols in the phonetic transcriptions are mapped on control codes for a commercial speech synthesizer. The control codes are sent to the synthesizer where they trigger the production of synthetic speech.