the context is that i found a pdf with what is probably persian and when i copied the text off of it its encoding got messed up, but kagi managed to fix it (i was suggested that maybe their llm is encoding-agnostic?)

@rebane2001 I had a situation where I was using Whisper STT to transcribe german speech, and pretty much early it started hallucinating, repeating the same word over and over. I threw this text into Kagi Translate, and somehow translation matched the original text.