P: Show words omitted from prompt or violates content guidelines

[stefaniek51216624]

Tried to make a word in colorful teddy plush but it says me it is against the guidelines. 

So I tried to say fur instead. That worked but didn't give me the result I wanted. 

[defaultap09zcxqfx55]

yes, I'm a bit embarrassed by my typos, but it still was worth showing how little Adobe cares, not to mention how little they pay attention.  Hey you just said you will no longer support our company, do you want to buy something else from us?

[defaultap09zcxqfx55]

Unfortunately , in addition to their canned response "our goal is to provide a safe ......" Adobe has already come right out and said that they refuse to tell us what the restricted words are and that they will not do so in the future. Their reason being that if we knew the restricted words that we would find other ways to say it in order to get the results that we want. $60 a month subscription, another 50 if I wanted substance 3d, and of course I'm the silly one because I think I should get the results I want instead of the ones Adobe says I'm allowed to have.

[@mj]

Here's a strange one. My initial prompt generated a result. Then I chose one of the options I liked and tried to create additional gens based on that one.

I got the Oops monkey man saying I was being naughty. 😞

Happen to anybody else? Do we know why that might happen?

Best

mj

[Javier37626234p7x6]

I got censored typing "bread". Adobe should not do anything other than implement age verification. Incredible the power of generative AI that does not draw bread so as not to hurt our sensitivity. My 5 year old son draws bread and if you ask him he will surely draw you a knife, but Adobe Firefly doesn't do it to protect us... Thanks Adobe

[droopydog500]

Hi @Javier37626234p7x6,

Can you please share your full prompt or a screen shot?  The word "bread" is not blocked for me:

full loaf of white bread 

Thanks,

    droopy

[Matthew364971072m1c]

one of the MANY problems with adobe AI,  not only are they incredibly restrictive, but they are extremely inconsistent .

[Rishi Ed]

Please provide a list of "bad" words that Adobe doesn't like. The list must be extensive because even something as simple as "Dumb people reading" gets rejected. Countless other examples have occurred. I have to craft work around descriptions almost every time I use Firefly. Regarding the mentioned prompt, I was hoping for something cartoonish and funny. AI has no sense of humor. Please fix.

[jane-e]

Rishi Ed wrote: even something as simple as "Dumb people reading" gets rejected

I'm glad to hear that Firefly rejected the word "dumb", as it gives the message that a disability makes a person flawed or “less than” — that disability is bad, negative, or something that needs to be fixed.

See this post from the Harvard Business Review if you are interested in learning more:

https://hbr.org/2020/12/why-you-need-to-stop-using-these-words-and-phrases

Jane

[SkepticOne]

Let's face it, Firefly has a lot of problems including those surrounding their "woke" programmers who decide certain words have the connotation THEY use. One wonders if they even speak English and realize many words have more than one meaning. I am frequently dissatisfied with results when I ask for artwork in a certain style or to imitate a certain artist. Cannot do people at all, inevitably when asking for blond hair the skin color and features are hardly European are dark to black. Firefly is far behind other AI generators...too bad, really. It would be useful for those of us who use Adobe products like PS and AI.

[Rishi Ed]

Not sure who you represent - Jane. I hope it's not adobe. I grew up with
"Dick and Jane" books. Perhaps you are also aware of the studies that some
names are positioned for failure?

It's a freedom of speech issue. Honestly, I don't care what Adobe "bad
words" are. I am asking for their list to decrease "mistakes" in prompts.

[self-promotion web links removed by moderator]

Thank you.

[Steven33847756abb6]

You are absolutely right. When I try text to image generative A.I., I can't
even say "sexy woman" without it being rejected. It is also hopelessly
P.C. I asked for brick layers for the 1666 London Fire restoration, and
they had one black man, one Asian woman and one white woman laying bricks.
I think they have some kind of quota system. I would like to try to have it
do a group of sperm donors and see how many women they put in....except the
word 'sperm' wouldn't get past the censors.

*Sincerely yours,*

*Steve *

*Steven W Rodgers*

(weblinks removed by moderator)

[TheChunkyMonkey]

Just use Dall-E save yourself the headache. Stopped using woke Firefly months ago. However you can find work arounds to get it to work. I can get firefly to generate nude women so they not as smart as they think. 

[jane-e]

TheChunkyMonkey wrote: "Just wondering why Pepe the frog is a word that violates the guidelines? Is pepe a bad word? "

Here are the user guidelines for using Adobe Generative AI:

https://www.adobe.com/legal/licenses-terms/adobe-gen-ai-user-guidelines.html

Note Section 4:

4. Be Respectful of Third-Party Rights

Using Adobe’s generative AI features to create, upload, or share content that violates third-party copyright, trademark, privacy, publicity, or other rights is prohibited. This may include, but is not limited to, entering text prompts designed to generate copyrighted, trademarked, or otherwise infringing content, uploading an input or reference image that includes a third party’s copyrighted content, generating text that plagiarizes third-party content, or using a third party’s personal information in violation of their privacy or data protection rights. If you’re not sure whether your content violates the rights of a third party, you may want to reach out to an attorney or consult publicly available reference materials at the following:

• U.S. Copyright Office
• U.S. Patent & Trademark Office
• Lumen

Pepe the Frog is not allowed because it is copyrighted by Matt Furie, the creator.

"In June 2019, Furie received a $15,000 out of court settlement in a copyright infringement case against Infowars and Alex Jones concerning unlicensed use of the image of Pepe the Frog on far-right themed posters. Furie stated that he would continue to 'enforce his copyrights aggressively to make sure nobody else is profiting off associating Pepe the Frog with hateful imagery.' "

https://en.wikipedia.org/wiki/Pepe_the_Frog

Jane

[drpepperandmilk]

Instead of just flagging words, flag contextual meaning of the words. AI should be able to determine from a prompt that "shooting" in the context of a "photograph" or "bleeding" in the context of "colors" are not a problem.

[@mj]

Agreed

[crimzoid27]

The content restriction algorithm in Firefly desperately needs some tweaks, because half the time it's unusable. It wouldn't even let me show someone kneeling in a completely innocent context. I have to go elsewhere for AI generated images most of the time.

[junction125]

I am sick of the monkey telling me I am using words AI won't allow. You can use "George Washington" but you can't use "Donald Trump"? 

[Kartika Rawat]

Hi junction125!

Thanks for writing in. Please review user guidelines here: https://adobe.ly/3DIxxLR

Let us know if you have any questions.

KR

[aqsa_sana8472]

"Yeah, it’s pretty frustrating when AI moderation feels inconsistent or arbitrary. It's like there’s a secret rulebook we're not allowed to see. I get the need for some guidelines, but it'd be nice if they were transparent and made sense across the board!"

[Orlando294138771lu0]

This is my biggest complaint in 2024. I volunteered for a UX interview with the firefly team a year ago, at which point I thought that controls were too loose and it was too easy to generate vulgar, nude, and inappropriate content through image generators.  Firefly has way overcorrected. Using the 'Generative Fill' tool, just 5 minutes ago I had the following 4 prompts rejected based on user guidelines:

1. A Gauntlet remiscent of those worn by characters in dark fantasy movies.

2. A bracelet, the type a vampire hunter would wear

3. An evil halo

4. seaweed hair

it can really feel like Adobe only feels comfortable with me prompting me sunflowers and and the happiest of terms. Unfortunately, I find the service more stifling by the day. 

[Anphonic]

Tell users what word was "removed". This prompt shouldn't have anything that's trademarked, copyrighted, etc

"I want an image that appears to be 2 photographs side by side, of an in ear measuring thermometer. The left should read 99. 4 and the right should read 100.7. The meme styled text is "When your low grade fever graduates ". Both left and right sides of the image should be essentially the same with the main difference being the temperature"

[Oh.N8]

Hey @Anphonic 

You may find generating the images seperately could work. once you have your images. Use Adobe Express. In Adobe Express you can place images side by side. 

A quick tip keep your prompts specific and short.

In terms of getting a specific number and thermometer, create an image (a drawing for example). Make sure the image is large enough for FF to read. consider using bold numbers in the thermometer reading. 

Cheers

O.Nate

[nix_2221]

"may be considered too strict for some users" lol, this waste of AI potential flages EVERYTHING; nine out of ten prompts are rejected, no matter how harmless and polite, any prompts longer than few words is rejected, i have to rephrase it and rephrase it and rephrase it like ten times before it miraculously does not flag it altho there is not a single inappropriate letter in it. It is completely and uterlly useles, what a waste, i would never pay for this service like it is, i would not use it even if it was totally free, it is that useless.

[Rubber Johnny]

NIX:  I hear ya, and I feel the same.  Firefly is garbage, the way they censor/flag so many prompts...to the point of uselessness.  Adobe clearly isn't embracing the AI tech as they should...which will hurt them in the present and future.  
They'll get left back eventually.  I guess I should be fine with that, as much as Adobe has pissed me off with other things in the past, such as their subscription models and lack of addressing many bugs.  Maybe I should instead be celebrating their inevitable demise.  

[Ben_Willemsen_6711]

I often have no idea why a prompt is not executed and can be marked for evaluation. Even after editing the prompt, it continues to be marked. I am completely in the dark about what I am doing wrong and have no idea that I am breaking the rules. So mark or highlight the word(s) or part of the sentence that are involved.