Generative AI is crazy powerful, but let's not pretend the limits don't exist.

Generative AI is legit powerful now, but just hyping it as all-mighty is just as irresponsible as pretending it has no flaws. It still can’t keep character consistency across images, long text layouts mess up all the time, and anything with precise physical structures like fingers or mechanical joints tends to break.

Not to mention the unresolved old issues like copyright and bias baked into the training data.

Admitting this isn’t hating—it’s knowing where the limits are so you can actually use it where it shines. The stronger the tool, the more the user needs to stay sharp.

I still haven’t found a solid solution for character consistency across images, it’s driving me nuts.

Long text formatting is a total disaster zone, text on signs messes up 8 out of 10 times.

The mechanical joints are even more messed up than the hands, and the gears just fused together.

“Admitting the limits isn’t the same as doomposting” - well said. People using it need to stay grounded.

Copyright and bias, those two old issues, people keep dancing around them.

Honestly, hands are way better now. It’s teeth and ropes that tend to mess up instead.

The stronger the tools get, the more you gotta keep a clear head. So many people hype them up like they’re some kind of cure-all.

When I’m doing commercial work, consistency is my biggest nightmare—the character’s face changes the second you switch angles.

Honestly, it all comes down to the training data when it comes to bias. There’s no way that’s getting fixed anytime soon.

Bump