Fifty-three subjects evaluated the naturalness and acceptability of three text-to-speech systems in French which are Loquax, Televox, and Multivox. Their evaluation bore upon 25 5-point perceptual scales. The main results indicate that the performance profiles of the three systems are very similar to each other. They also reveal that the average evaluations of the three systems over the 25 scales are negative (Multivox: -0.14; Televox: -0.19, Loquax: -0.38) and do not appear to differ significantly from each other. An interesting result is that the oldest system, Loquax, received the largest number of -2 ratings from the subjects, whereas Multivox obtained the largest number of +2 ratings. Some research directions are proposed for the future.