This paper describes a Spanish Syn thesis-by-Rule system, which is completely simula ted on a PDP 11/60 minicomputer and produces an intelligible and natural speech. Input to the system is any orthographic text. A linguistic preprocessing converts this input to a phonemic text with stress and pause marks included. Then a phonetical process inserts prosodic parameters and evaluates the acoustic parameters of each phonetic segment, wich are then used to feed a vocal tract model, that generates the sampled waveform, ready to be outputted to a D/A converter and a speaker. The vocal tract model wich has been choosen is the formant synthesizer proposed by D.H. Klatt.Automatic letter to phoneme transcription is based on rules stated by the Spanish phonologist A. Quilis. Besides, the set of algorithms for phoneme to parameter conversion has been extracted from an extensive experimental and original work carried by the authors on acoustic phonetic characterization of Spanish sounds in their spectral and temporal behavior. (A cassette recording of speech output from the system will be played).