Comment by MattRix
13 hours ago
To be clear: they already had image generation in ChatGPT, but this was a MUCH better one than what they had previously. Even for you with your stable diffusion app, it would be a significant upgrade. Not just because of image quality, but because it can actually generate coherent images and follow instructions.
As impressive as it is, for some uses it still is worse than a local SD model. It will refuse to generate named anime characters (because of copyright, or because it just doesn't know them, even not particularly obscure ones) for example. Or obviously anything even remotely spicy. As someone who mostly uses image generation to amuse myself (and not to post it, where copyright might matter) it's honestly somewhat disappointing. But I don't expect any of the major AI companies to release anything without excessive guardrails.