File: f09e1c26-1767-48fd-a398-a(...).png (2.5 MB, 1024x1536)

2.5 MB PNG

Certainly made him fatter than NBP does

File: 1765931823093-019b29bc-8f(...).png (1.74 MB, 1024x1024)

1.74 MB PNG

>Banned for (((racism)))
>Obviously everyone else's racist posts remain standing.
Classy as always and never change, 4chan.
>>107575271
lmarena. It doesn't look like it's running at lower quality there atm, but I'd be careful because 4o/GPT-Image-1 is definitely sereved at low quality settings there.

It's kind of curious that they didn't mention anything about Sora, anyway.

File: 1744716498441376.png (1.87 MB, 1024x1024)

1.87 MB PNG

>>107575523
>lmarena
thx
model is eh
better than gpt-1 but not as flexible as nbp

File: gpt top mj bottom.jpg (545 KB, 1024x2383)

545 KB JPG

>A cinematic 1980s Christmas photograph in color, a large, beautifully decorated Christmas tree dominating the scene, warm glowing string lights and ornaments filling the space, a tall, slim Korean man and woman dancing and spinning near the tree, their figures slightly smaller in the frame, long slender silhouettes, elegant movement, both wearing cozy knit sweaters, natural motion blur capturing joyful movement, intimate, candid, music-filled atmosphere, vintage holiday color tones, warm reds, soft greens, muted gold highlights, slightly faded colors, analog film look, gentle grain, soft contrast, editorial photography style, timeless and nostalgic, vertical portrait composition, 3:4 aspect ratio, album cover and Instagram post ready, no modern furniture, no contemporary styling,

still mogged by mj
I honestly don't they'll ever match the visuals

File: Clipboard.jpg (1.11 MB, 2048x1536)

1.11 MB JPG

>>107575651
Considering how they're offering it at lower price than GPT-Image-1 per picture on API and how it's relatively fast, it's probably quite optimized, compared to what they *could* do. I might like NBP better, especially in terms of imitating, or trying to imitate, specific styles. But I guess it might be good for some things.

You could want to check out the model on Sora, too. It's different, for sure. Sora gives 12 gens, at least.

Gotta say that I really kind of preferred the more "natural" looks of women in 4o. This is the same prompt in hybrid painterly-anime style about a "zaftig 25-year-old Neapolitan woman in a bikini"

File: 1765935517307-019b29f4-e1(...).png (1.84 MB, 1024x1024)

1.84 MB PNG

Living artist names remain as blocked as usual, it goes without saying.

File: gpt vs mj 23h anthropomo(...).jpg (599 KB, 1024x2560)

599 KB JPG

>>107575765
at least they toned down the piss filter

File: 1736469124148888.png (2.79 MB, 1024x1536)

2.79 MB PNG

1.5 reminds me of the first versions of SD

File: gpt-image-1.5 try on clothes.png (1.17 MB, 2048x832)

1.17 MB PNG

kek, you have to ask to keep the tiddies

With the new sora model forcing me to redo all my prompts, anyone know which AI would be good for trying to turn an image with a specific art style into a prompt?

File: 1755125190082265.png (15 KB, 346x346)

15 KB PNG

>>107577612
https://huggingface.co/spaces/fancyfeast/joy-caption-beta-one

with most Extra Options enabled

>>107577828
Still does pretty bad at picking out the fine details of that one weird mutant from the old RPG.

Maybe I should just make a collage and see if that helps

File: 1763734657234882.png (12 KB, 336x518)

12 KB PNG

>>107577974
check Caption Length to very long as well, and try the different Caption Types (SD, MJ)

>>107577986
Trying it out, just seems outdated when I can put a whole novel into a Sora prompt.