Just to be clear right off the bat, I DESPISE this change.
Pros:
Generated images can be very realistic and of high quality.
New image size is available.
Cons:
Generates different people with different facial features every time, there's no consistency, no matter how many different prompts or entire sentences were provided.
Even after trying to avoid weird, deformed images one'll get a few straight-up UGLY AS SIN or absurd images in between a number of generations.
I swear I'd never get anything even close to being "ugly" with the previous model, unless I prompted something like 'Cursed photo'.
The model can't draw any resemblance to celeb faces anymore, even if one tries, it generates a random face.
The amount of time it takes to generate now is around 3x longer, sometimes it just fails and retries.
Lastly, the previous model, which had been in use for over a year, was replaced overnight, like literally. All the prompts that everyone had created over so long and with care, suddenly became meaningless. I had to create an account just to check if it's only my pain that I've to bear.
There is consistency if you use the same seed to generate images. I frequently save the seeds so I can create several variations of the same picture. In this example, I have the same woman being generated, sometimes with different glasses or minor variations:
Haven't had this issue. I've had it far more with the old model than this one. Especially messed up faces when the subject was slightly distant.
Some people have posted some celebrities on reddit, like Taylor Swift. But yes, far more people found many celebrities that didn't work.
This one I agree. Seems at least better today, not sure if something was changed at all.
It's a stretched solution and does not count as a positive side of this model either. Actually, you'd know the issue in the first place what I meant by not creating 'consistent faces' if you had been working with realistic faces for a year or so, which I doubt