Nudity and other issues which appear to violate Adobe Generative AI Guidelines [merged thread]

For whatever good it might do, when gen fill gives me an incorrect violation warning, I take advantage of the option to submit an example of what I was trying to do, along with the image in question. With luck, Adobe programmers are taking such errors into account.

The problem is many of the time it blocks things, that are not violating the TOS. Family vacation photos are blocked if someone is in a bathing suit. Professional model photographers can hardly use generative fill for touch-ups and corrections. Yes, you can use the old ways, but we are paying for the updated technology to speed up workflow, as Adobe has described their latest products. generative fill blocks things that have too much skin. When trying to edit fingers or toes it blocks constantly. It's not even nude content and it's blocking non-violating things.

Bottom line: If it's not breaking the TOS, it shouldn't be blocked. Otherwise, update the TOS to correctly reflect the filters.

I get around it in Photoshop itself. It is fairly easily tricked. As long as it can't see a feminine-looking person in the photo, it'll let you do background removal, canvas extend etc. If I want to do canvas extend (my most common use because I get photos that end up needing to be 1:1 cropped for socials but the subject is cropped too close already), then I crop out most of the woman, extend what remains of the canvas, copy that new image, CTRL+Z to get the subject back, change canvas size and drop in the extra. It's frustrating that Photoshop won't let us do it directly, but we can get around it, so that in the end their restriction is functionally and totally pointless at achieving whatever weird aim it is trying to get at.

[and on another note, I recently paid for Luminar Neo's Gen AI features and the restrictions are essentially identical -- if doesn't work in PS it won't work in Neo either]

@Nancy OShea 

I agree that "it won't be resolved in this public discussion."

The current situation seems to be that artists feel this: "for so long, we've used Photoshop to handle these types of photos or images. Now they release Firefly within the Photoshop ecosystem, we naturally assume it should handle the same tasks it always could, right?" And unless explicitly blocked, it should indeed have this capability.

However, artists find that's not the case. Firefly has imposed certain restrictions, resulting in an inconsistent coverage of functions within the same ecosystem. Furthermore, it over blocks some uses that weren’t even officially claimed to be blocked.

For some artists, this feels somewhat akin to the transition from traditional hand-drawing to digital drawing, where they might have expected digital tools to do everything hand-drawing could. But instead, the digital product says something like, "The night is evil, so we won’t give you black color to draw." This will make them unable to even draw black hair and feel frustrated.

So yes, artists complain here won’t change much, but it’s understandable why they want to complain. After all, the advertisements used to entice users to pay for the service didn’t mention these functional discrepancies.

Currently, artists looking for change are left with a few options:

1. Directly petition Photoshop’s official team in hopes that enough collective feedback will lead to changes.
2. Hope for a less restrictive competitor to emerge, allowing them to migrate.
3. Use a more cumbersome multi-stage process, such as exporting from Photoshop -> processing via Tensor.art (or another unrestricted external AI tool) -> returning to Photoshop for final edits, thereby bypassing Firefly’s limitations.

Good artists know how to create art with whatever tools & techniques are available. They've been doing it for millennia without any assistance from computers or open AI tech.

That said, AI can be a useful addition to the artist's tool box, but it's not a one size-fits-all solution for everything. Knowing when & where to use AI effectively & responsibly is the challenge we're all still coming to terms with. And it won't be resolved in this public discussion.

Why do you think you need AI to combine (composite) two images?

By @Nancy OShea 

This is a very funny question, like...
Why do you think you need Photoshop to draw?
Why do you think Leonardo da Vinci needs oil paints to paint "Monna Lisa"?
Yeah, there are many ways to do things, artists just "can" pick any one way they want. Pencil, oil paint, fingers, scissors and paper, camera, graphics tablet...
The reason can be "it's more creative", "it's faster", "it's convenient", or just "they want to try it".

Keine Einschränkungen 

@QINGCHARLES I can help you, had the same frustrations but found a good workflow solution inside of photoshop.

"...we should just use Sensei. I looked that up because I didn't know [what] it is, apparently it only works on Premiere and AE..."

By @QINGCHARLES

==========

Photoshop has been powered with Sensei technology since 2021. See below for more details.

Sensei is the name of older AI powered tools in Photoshop. Content aware fill and repair brushes.