A signal-domain approach for controlling intonation and stress in segment concatenated speech is reported. The extraction and editing of pitch-period in the signal is shown to be an effective way to deal with these suprasegmentals. In addition to the generation of querry sentences, generation of songs is also reported.