Conclusion

Although the list of current applications for speech synthesis systems is extensive, it would be wrong to assume that speech synthesis is a fully mastered technique. However, years of research have enabled us to achieve a quality approaching that of the natural voice.

Quality of synthesis is a crucial issue. It is essentially manifested by the fact that understanding synthetic speech requires a greater effort on the part of the listener than natural speech. This extra effort is made necessary by possible processing artifacts, pronunciation errors, insufficiently expressive prosody, or more generally, by the unnaturalness of the elocution. At present, while the quality of synthesis systems based on the selection of non-uniform units is sufficient...