Bug Reports

"closeup of dragon scales with gold wire"Has produced "We encountered a safety issue with this image. Please try again." error without other explanation. Repeating this prompt generates 1-2 image blocks with this warning.

I hestitate to call this a "bug" because it's more of a very minor UI thing, but I guess close enough. Either way, this definitely isn't anything critical, but figure I'd point it out. For Context: In most cases a style will be applied to the generation, which means that below the prompt text box, it shows a list of the styles and a "clear styles" button. I'll attach an image to show what I mean.But when you have no style / content type / color etc, that bottom row of style tags (below the prompt text) is not showing, which causes the four image results to resize ever-so slightly (because they have more room on the page and expand). Problem Scenario: If you want to switch between the different 'Content Type' options to compare which version of a result you like, when going from no-style to any-style, the images do a little animation and resize. Which makes it a bit harder to directly compare a style to the none-style. In other cases, where all the images stay exactly the same

I took a Firefly image into Lightroom Classic and added a preset.  Upon export, I made sure Use all Metadata was checked on.  When I uploaded the image to Content Authenticity Verify, the credentials were missing.  When I opened the same image in Photoshop from Lightroom Classic and exported it, the credentials were intact. Just info for the Adobe team.

Small bug noticed frequently when testing generation of text:There are elements (probably intended to be placed within text bounds) that fall off into the negative space.--I haven't been able to see what triggers it most... perhaps there is a trend?  

Hi there, I don't know if this would qualify as a bug, however its worth pointing out. I am generating an image using the following prompt: "Velen Swamp Crone witch ugly decay mottled sickly mystical magic legend baba yaga" And I receive the Uh Oh popup that a word was removed during generation  Since it doesn't tell me what word was removed, I tried downloading one of the images and unfortunately, my entire prompt (including the mystery violation word) is used as the filename. Firefly_Velen+Swamp Crone witch ugly decay mottled sickly mystical magic legend baba yaga_art,dramatic_light,blurry_background,hyper_realistic,bokeh,fisheye,chaotic_12286.jpg This is what I'm reporting as a bug, because if the word is removed from the prompt, then it's not going to match what was used to actually create the image. It would be great however if the UI told us what word is in violation (like highlight it or something?) - I feel like I'm playing a guessing gam

I had "Redwood forest exploration" as my main keywords and wanted to add a couple hikers into the illustration but it spat out an error, "This image produced an error. Please try again." Perhaps it is a safeguard from the AI feeding inappropriate imagery but I wish the messaging was clearer as to why it was not working. I tried "hikers in redwood forest" and "Redwood forest explorations with hikers" but all variations produced the same error.

How is the word "Trump" a violation of your terms - he was the 45th president and might be the 47th. The anti Trump bias is gross and offensive.

Dear sirs,Thank you for the early access to the Beta version of Firefly. I just wanted to report a bug. Maybe it's just the server, or maybe my browser, but the text effects is working. I get this message every time (see screen shot)Best regards,Adrian Sommeling 

I get this error even though I have internet access

When I click on the text to image prompts in the homepage and in the gallery- they all go to text effects. safari Mac version

Germany's Capital "BERLIN" isnt' recognized as a valid word.It works with Helsinki, Tokyo, New York City, Moscow but not Berlin.

I am a daily user of stable diffusion, midjourney and DALL-E. When I notice that I ask for certain picture: "picture of a streetview decently filled with people in Amsterdam taken at night, small street, people walking, Food shops, bars, pubs, cobblestone road, masterpiece, beautiful lighting". I notice that the people in the picture kind of look like persons but you can see they are vague lines that resemble people. This may be because they are a bit more far off in the picture? But there is a similar issue to other object like bicycles, chairs, etc. 

I tried rendering some images and they would not populate.

It's understandable that text changes aren't immediately reflected. However, when I change the text and then update another parameter it renders new compositions but ignores the text change. I erroneously submitted a downvote on a result thinking it was ignoring my text changes but it seems they weren't sent. I found this behavior in both text to image and text effects.

Lord Shiva's face is distorted. Keyword - "Lord Shiva in space"