Is it possible to use UTF8 for CLAM speech? I tried the sendtoaddress command (after fixing clamd, pull request issued) on an UTF8 terminal, so clamd should get the UTF8 string. I used this string as the CLAM speech parameter "test2 äöü 象形字", transaction: 2e7d31559f88322bbd66ea361f312c695fb62984ae133b7661cd21dcb2110366. I don't know if the UTF8 characters made it into the transaction, looks like khashier.com doesn't show it, but it looks like the typical 2 bytes UTF8 chars.
Update: khashier supports already UTF8, test with the text 嗨!:
http://khashier.com/tx/e4d5c2f48e8da1c0ed800edf87281dddb9b8d1c5e612e18b173015c7946707b0Looks like some kind of command line encoding problem on my Linux machine or in the clamd command line program.
Not sure if its applicable, but is it possible to give a file as an arg for you clamd clamspeech? If so, that might avoid any tinkering by the shell/locale.