Interestng. I agree that the Parti example is better/more accurate than both Dall-E2 and Midjourney. I think what's perhaps most impressive is how well it captured the description. The Dall-E2 version looks fantastic, but the tie is less "colourful" and it's not as clear that the beaver is in a library—almost looks more like an office to me (possibly the thinness of all the spines, and lack of Dewey labels). The Parti version's face looks a little more illustrated than "dslr", to me, but that could partly be because it's smaller and thus has fewer pixels for detail. I also find it weird that the stack of books all seem to be the same book, or at least look very similar. Checking out an entire encyclopedia maybe? Heh...