Those old speech synths were mostly getting fed phonetic data, which is an altogether easier proposition than true TTS.

Rob