Hi jairoaf,
Yes, the automatic detection uses the start of the audio and sometimes it might guess incorrectly.
I am preparing a new version that will allow the user to select the language manually as well, so it should fix this issue.
