OpenAI is also releasing DALLE-3 in "early October" and the images they chose for their demos show it demonstrating unprecedented levels of prompt understanding, including embedding full sentences of text in an output image.
Not unprecedented at all. SDXL Images look better than the examples for DALLE-3 and SDXL has a massive tool ecosystem of things like controlnet, Lora’s, regional prompting that is simply not there with DALLE-3
