Hello, Indeed, gespeaker doesn’t find the mbrola voices since it requires espeak data to handle them and looks for them in the wrong place. The path where it looks for the espeak data seems to be hard-coded in the source, namely in /usr/share/espeak-data while the espeak-data package installs its data in an architecture-dependent directory (/usr/lib/x86_64-linux-gnu/espeak-data in amd64, etc.).
This means that a symlink ln -s /usr/lib/x86_64-linux-gnu/espeak-data /usr/share/espeak-data provides a really dirty workaround. I imagine a proper bug fix would be, in src/EspeakFrontend.py function loadMbrolaVoices, to build the variable pathVoices from the host architecture. Best regards Sam