The current implementation reads from the initial list when pressing `s` to speak. Fix that.
The current implementation reads from the initial list when pressing
sto speak. Fix that.