P: Generated images violate user guidelines

Yeah, I ran into a similar roadblock while using the generative fill feature on a clothed photo of me flexing my biceps (I'm a bodybuilder) and received the same error. But I stumbled upon a workaround that could tide you over until Adobe rectifies this glitch. Here's what you need to do: pinpoint a part of the image that's not sexually explicit (or that could potentially be misinterpreted as such), and form a new duplicate layer from this unexplicit area. After that, select this newly created layer and use generative fill on it. So, if you're trying to modify the mouth, for instance, just make a separate layer of the face, select the mouth, and use generative fill on it. I found this technique effective. You can also select everything that is explicit, invert the selection to everything that is not explicit, and create a duplicate layer of that.

That being said, I'm pretty annoyed at Adobe's decision. It doesn't make much sense, especially when a lot of us Adobe users (myself included) often work with stuff that can be seen as sexually explicit. Some folks even use the software to edit proper adult content. I don't get why this would be a problem. The AI certainly doesn't mind.

I guess the only reason that makes sense is that this wasn't really Adobe's own choice, but a decision they had to follow. It's kind of like how ChatGPT doesn't write sexually explicit stuff; the limits probably come from the people who built the AI that Adobe uses to recognize patterns in pictures. They might have put in rules to stop people from using it with sexually explicit content. Even though this still seems like a weird decision by the developers, it suggests that Adobe might not be the one to blame for this.

All I've been using it for is on mock-up images of a plain wall and a picture frame. Sometimes I'll resize the picture frame and to smooth it out I'll use generative ai. It works a lot of the time, but also a lot of the time it says the generated image is prohibited. It's literally a plain wooden picture frame, with nothing in the picture,  on a white wall. Scandalous 🤣

This has happened on several occasions and is becoming frustrating. 

I make a selection inside an image, go to generative fill and hit submit. (no prompts input) 

I want to remove an object from a picture. Then I get a notice stating this violates community policy blah blah blah. 

I have done this before with no problem. Why all of a sudden would it violate anything? This continues to happen. Very frustrating. 

I posted this in another thread a couple of days ago, but not sure if it was the right one.   THIS is what CAN be done using the new AI tool right now !  This is a musical video I made up using entirely generated images from the AI. I added a few bits and bobs  but so far the comments have been very satisfying. I recorded the music myself playing my Yamaha Tyros 5 keyboard and mixed it all using Captracks... I think it's nice to give credit to the team for a change instead of whinning all the time.

https://youtu.be/7pt0Wn3xKA0

When I get the dreaded orange error, I usually type in remove object. Works for me every time 

I experienced the same issue on several pictures of antique pens. Adding a prompt solved the issue. I've been trying to upload an image to the dropbox folder but it always stalls at 10% uploaded. It's a picture of an antique pen on parchment. I think it mistook it for something else. I added the prompt "fill in the style of the image" (probably more than I needed). All good. Thanks for the solution.

And - I agree with generating locally. I'm traveling and have been in an area where wifi and cell coverage are minimal and intermittent. I spent a lot of time this morning trying get get a couple of very simple fills done. Not critical work, but telling if you're in poorly connected area. I have SD/Automatic1111 running and it clicks along nicely. But, this photoshop featur is awesome! It's right where I need it to be - in the workflow.

Cheers,

@williamh53049808 its a beta. Most users are getting warnings by not adding a prompt and leaving the field blank.

Havn't seen the guidelines either

I agree, I am running Stable Diffusion locally. Yeah it is not nearly as fast, but at least its local. What I am wondering is where can I read the guidelines? I don't remember seeing any. I just loaded it up for the first time today, and tried extending an anime picture, obviously sexual but not nude (i was just trying it out to test the feature out), it extended it no issues and I got no warnings (no text prompts were used). From the sound of it, i should have. So it sounds as if the warning system is not consistant. 

Daft thing is, if you type in posing, posed, poser, or pose in the prompt box, 90 % of the images are featuring scantily clad ladies, which adds to the irony of it