In conclusion for these images above:

The word filter was never activated, but many image results were blocked after finishing. Wrestling was almost impossible to get an image for, and there is a general fear of all females in sports. Part of this may be because the end-result filter also blocks "visual schizophrenia" - many of the images that came through still had this, such as random body parts or deformed anatomy. It's not possible to tell if the body intense sports like wrestling actually are just too messy image-wise for the AI to generate images from it, or if it fears the nudity. It however does not have any problem showing bare chested men.

The results for "boxing" are outrageous.

Note:

Swimmers are banned, swimsuits are banned, boxing and boxers are banned, wrestling is mostly banned.

But they're not actually banned if you rephrase the prompt. Very silly.

A more complex prompt for trying to create a representation of the Metal Slime Voodo materialist.

This wasn't easy at all, not because the images are blocked, but because they just turn out weird. And stylistic in a specific way no matter what.

>>7651
Also that's a lot of bare skin. It seem whatever filter they have doesn't react to dark skin, thinking it's clothing.

Or, Idk. Some of those above turned similar but only randomly.

>>7652
Nvm. Maybe it's this prompt. Removed "african" and you can literally see her buttocks in one of these.

They really need to fix this so normal prompts create normal images and not softcore porn - which then is blocked after finishing.

All because the AI is thinking in lewds all the time and then regrets it right after.

TRIGGER WARNING

No further comment.

I guess the filter actually does just remove visual schizophrenia aside from what it perceives as direct attempts at breaking the TOS. It's the interpretation of innocent prompts as not-innocent that's the issue here.

These images are just on the edge of containing random body parts because the image is too complex for an AI model with no internal perception of 3D shapes.

Often it's obstinate to the max, but for this theme it instantly got what I meant even if I only described the image. That even looks like the EU HQ in Brüssels ;PP

While the images look fine in themselves, why is it so hard to understand that a vickers team has an actual vickers machine gun (and not some other weird contraption, omg).

A + for propagandistic comedy though ;P

Asking Bing to create an image based off a sketch.

I saw the Boeing space shuttle had a "gas leak". Why not ditch space shuttles altogether?

Really pushing the limit here.

Ok, I know.

It's my own "fault" for trying to create these, but this is legit scary. Some of them were so intense I didn't even want to save them.

>>8027
Best viewed in full size.

It either recognizes art, or the censor algorithm just broke here. Overall it does pretty well with religious themes anyway.

It seems there has been an improvement in the accuracy of dall-e3 via Bing, but since this can't be in the image generator itself(which hasn't been updated to my knowing), they must have improved the layer in between, the way chatGPT is used to generate middle-prompts based on what prompt the user inputs.

Pic related was impossible to create just a month ago. It would just downright refuse to show any result (the prompt being innocent, the image was just not shown).