Before anything is saved, each character Tessitura can’t store is swapped for its nearest equivalent. Accents come off only where they have to, so Dvořák becomes Dvorák and keeps its á. Other alphabets are spelled in English letters, and emoji become the shortcodes people type in Slack. Left alone, these characters are changed or lost on the way into Tessitura without anyone being told.
The swap is made by any-ascii, an open-source library that covers practically all of Unicode and runs in .NET, PHP, Python, Java and JavaScript, so the server can make exactly the change this page describes.
Romanising other alphabets is a guess at best. It reads Japanese kanji as Chinese, and spells Κάλλας “Kallas” where she spelled it “Callas”, which is why the message asks people to type their own spelling instead.
Tessitura keeps most text in varchar columns, which hold
one 256-character set instead of all of Unicode. On SQL Server’s
default collation that set is Windows-1252, which is Latin-1 plus
curly quotes, dashes, the euro sign and a few Central European
letters such as Š and Ž. Switch to Latin-1 to see the stricter limit.