You do not need to convert externally. You can use our Import UDF model and integrate this as part of the dataflow.
Handling multiple charsets can be tricky. Some of this may involve honing your "data cleansing" algorithm that could detect new charsets over time and apply the appropriate decodings. For instance:
win = line.decode('windows-1252').split(',') #windows-1252
norm = line.decode('utf-8', 'ignore').split(',')
ascii = line.decode('ascii', "ignore").split(',')
ascii2 = line.decode('ISO-8859-1').split(',')
So you can similarly look up your particular charset "latin-2" and do it similarly.
Could you post your code and some sample data, so we can check specifically what issues you are facing?