In this paper a new technique for coding speech signals for the use in high quality text-to-speech environment is presented. A prosodic control of the acoustic output is implemented merely by modifying the speech waveform during the resynthesis phase. Any degradation in sound quality is avoided, since no parametric representation of the speech signal is used for the storage of speech. This algorithm preserves the intrinsic features of the original speech signal (microprosody) while allowing for a prosodic control with only few values per time. The algorithm requires very small computational equipment for the synthesis.