I'm involved with a project, Rockbox, an open source firmware for several digital audio players. One feature we offer is voiced menus. We do this by pre-generating voice clips for each menu entry. Currently we pre-generate the voice using one of several open source text to speech engines. We discovered that we can use espeak to generate phonemes for the mbrola voices, and were unable to determine what specific license the output from your speaker falls under. We saw a posted license both for the program, and for the database, but it is unclear to us what we are allowed to do with the actual program output, should we record it to wav, convert to MP3, and wish to distribute to our users.