Repairing UTF-8 that had been stored as Latin-1. The correction is mechanical, not a guess: the stored text re-encodes to exactly these bytes. name: "Une Esclave Française" -> "Une Esclave Française"

Repairing character-encoding corruption (mojibake) introduced during data import; the person/ensemble name itself is unchanged. Cause: UTF-8 bytes were decoded as Windows-1252 at import, so each non-ASCII letter became a 2-character sequence. Stored value "Une Esclave Française" is being corrected to "Une Esclave Française". Detected by decoding the stored text back through the offending codec and confirming it yields valid UTF-8. Part of a sweep of all singers, conductors, composers, characters, operas and recordings; 419 fields matched.