Is it possible to have text output insted of Voice...
# feature-requests
k
Is it possible to have text output insted of Voice ?