In this paper we asked four Brazilian and four French news announcers, both females and males, to read a text in their native language three times: in a neutral way, imitating broadcast news from memory and imitating the style of another female news announcer after listening to her talking about a different topic. From these 24 recordings, we analysed the proportion of initial stress in accentual phrases, three fundamental frequency (F0) statistical descriptors (median, standard deviation, range), reading and pause duration, as well as spectral emphasis, a correlate of vocal effort. Results pointed out that the main characteristics of the imitated broadcaster style in comparison with the neutral reading are: a higher proportion of initial stress, an increase of at least 2 semitones in F0 median, an increase in F0 range (for Brazilian news announcers, especially) and F0 standard deviation (for French news announcers, especially). In all cases, a higher value for spectral emphasis was found.