I've been messing around with OCR to extract the text that Claude puts out, then pumping that text into a Text to Speech, here is how it's like
https://files.catbox.moe/eznfd3.webm