This paper presents a brief introduction about
text-to-spcech (TTS) systems, its main structure and
alternatíves. In a more specific way, it was developed three
algorithms of a system that automatic reads numbers. Each
algorithm hás its own functions and its own way to
approach the problem. The algorithms have been
programmed using Matiab sofhvare. The audio signals have
been recorded and edited using Praat software. Finally a
perceptual evaluation was made on each algorithm and was
assigned a rating to each one. Generally, thc MOS gives a
very good levei ofclassification for the three algorithms.
