Mojibake Fixer

In short The "’" and "é" problem has one cause: text that was encoded as UTF-8 and then read as if it were Windows-1252 or Latin-1. Each multi-byte UTF-8 character gets interpreted as two or three separate single-byte characters, which is why one apostrophe becomes three symbols. Because the damage is a deterministic misreading rather than a loss, reversing it is exact — the original bytes are all still there, just being read through the wrong lens, so the repair is lossless and not guesswork. That also explains why text sometimes carries several layers of it: each time the file passes through another system that decodes it wrongly and re-encodes the result, another layer is added, and up to four are unwound here. The permanent fix is upstream rather than in the text. Declare UTF-8 everywhere — the HTTP header, the HTML meta charset, the database connection and the column collation — because the single most common source is a database connection that defaults to Latin-1 while the columns are UTF-8. Fixing the stored text without fixing the connection just means the next write is damaged the same way.