Skip to Content

Department of Linguistics

Synthesis-by-Rule (SBR)

Robert Mannell

In early TTS systems the generation of the acoustic parameters that were sent to the speech synthesiser was handled by an SBR module. Such a module would apply rules and stored data values to a sequence of phonemes, or allophones, to generate time-varying acoustic parameters. These time-varying parameters would include more steady-state allophone target values as well as changing values during phoneme-to-phoneme transitions. It would apply rules to generate both steady-state as well as dynamic acoustic parameters.

  • For information on how MU-Talk's SBR branch works, click here.