[msgfilter] Invalid multibyte sequence when transliterating UTF-8 latin to cyrillic

Greets.

I am using msgfilter from the gettext tools to convert the msgstr's from
serbian latin to serbian cyrillic scripts. The PO file encoding is UTF-8.

Upon invoking the script
msgfilter -i input-file.po -o converted.po sed -f ~/work/lat-to-cir.sed

I get an error:
msgfilter: invalid multibyte sequence.

In the output file, all the characters in the msgstrs that should have
appeared in cyrillic are gone. No mention of the message in the docs or
on google in a related situation. Can someone explain what the message
means? For the record, the sed script lat-to-cir.sed is a sequence of
sed's "s"-commands and works as intended if invoked stand-alone.

f