When espeak loads a sound file, it checks if its characteristics (mono / 16bits / sample rate...) are in accordance with what is expected. If they are not, then espeak requires sox to convert the sample. Because sox is called through the "system" system call, espeak builds fine without the dependency on sox explicitly declared, however, espeak will fail converting the sample if media-sound/sox is not installed. So I suggest to explicitly declare sox as a dependency of espeak. ================== Reproducible: Always The call to sox can be seen as part of the file src/readclause.cpp within the function LoadSoundFile : sprintf(command,"sox \"%s\" -r %d -w -s -c1 %s %s\n", fname, samplerate, fname_temp, resample); if(system(command) == 0) ============ Of course this is not a regression. This problem was already there with previous versions (at least 1.44.05). I simply mention it with regard to the current stabilization request.
I am removing the block since this is not a regression.
This is fixed in espeak-1.46.01. Thanks for the report.
