Transcriber input strips text of a lot of nuances before the LLM gets it. Including normalizing the text when we don’t speak in proper English with proper punctuation