A possibility is, they make smaller pictures and use a up-scaling method. This would explain maybe the lost of details. It seams a sharpening algo is in use.
Maybe data reduction.
Surely wrong training data is used too.
Beside of the birdshit moon, the star night background is a other ugliness. before the stars are decent and a true background. now they are strongly over prominent, the same effect like with the moon.
Small faces in distance are more blurred and distorted.
Useful would be the support of seed and access to different models. It would allow prompt testing on a more professional level. But exactly this is what would reveal what they maybe want to hide.
I have not work with Bing, but it seam it was updated later. i have seen the quality change way before November 2024.
I have sadly not much moon pictures to find the exact update time on DallE under ChatGPT. But in case somebody has examples, we could find the update time for DallE like now for Bing.
This post shows it to most best way, with side by side comparison. August 2024. You can see among other things (moon, stars, nonsense backlight, nonsense back-shine, etc) the over sharpness of the new model. “This has been going on for 2 months now and they have been completely silent about it.”
if you can maybe post some side by side comparisons too?
I must say, i don’t like the communication policy of OpenAI.