This is quite a surprising model and probably the best I’ve tried. I often use it to read back my own work while writing. It’s a fantastic way to hear pace, cadence, etc, while also a really great way to find typos that aren’t misspellings. I do wish it handled em —dashes—. As it stands, anything inside an em dash is all compressed together into one single word. Which, isn’t very helpful and suddenly breaks the rhythm. Heck, I’d be okay if it simply treated em dashes as parenthetical. Also, sometimes the model will see the context of a character’s speech and punctuation, (eg: he shouted, “blah!” And the model reads it louder ) but often it ignores what the text is indicating. What would be even more helped if the model could for of scan the text around dialogue and try to infer what the dialogue should sound like, which is a big ask… short of that, if I could select some text (dialogue or otherwise) and have 4 or 5 options indicating how is should be read (eg. Quiet, thoughtful, concerned, scared, alarmed, sinister, etc) but only for that selection. Because the next paragraph might not need that inflection. Otherwise, this is a very cool tool. Impressive.