Back to the MaxMBROLA~ page
MBROLA voices contain phonetic informations extracted from a particular speaker. That informations are structured as a database of all possible diphones (segment of speech, which starts in the middle of the stable part of a phoneme and ends in the middle of the stable part of the next one) available in the language of the speaker. MBROLA databases only contain one specimen of each diphone.
All MBROLA voices can be found here (section "2) Getting the MBROLA Voices"). Then, the database (the file to be loaded inside the MaxMBROLA~ external object with the voice message) is the big file (around 8 Mo) without extension and having the same name as the package (for exemple: fr4, ca2, jp1, etc). The list of the phonemes available in the database (and the corresponding SAMPA notation) is provided in the text file which have the same name as the package (for exemple: fr4.txt, ca2.txt, jp1.txt, etc).