Firefly is getting worse

[Slobodan5FD7]

Why is it that firefly and adobe ai in general is getting worse and worse? it can't generate anything that even resembles reality.

[Kartika Rawat]

Hi Slobodan5FD7!

Sorry to hear that. Could you share a screenshot of the issue? Please share the prompt as well that you're using to generate images.

Let us know. Happy to help.

Thanks,

KR

[Rich_60]

I cannot agree more with Slodoban comment. I asked Firefly to insert a forest between my Hospital (reference uploaded) and the Lake Villarrica, to resemble reality. It just added a thin line of grass. My god!

[ndelcastillo]

I couldn't agree more. Whether you refine the prompt or styles, it just keeps pumping out almost identical images, and compared to other platforms, they're terrible. It used to be much better.

[Kartika Rawat]

Sorry to hear that. Could you share some prompts and screenshots of generated images?

Let us know. Happy to investigate

Thanks,
KR

[Oh.N8]

Hey @Slobodan5FD7 

 
Please remember to provide Feedback to the developers. Press the thumb up or down and follow the prompts.

As a user myself, I don't really see these images as a finished product. I like to add my own flare to the images in Photoshop: adding adjustment layers, expanding, and adding additional elments to the image.

Cheers,
O. Nate

[Offer-Readman]

100% agree. It basically ignores 90% of the prompt. It was definetly better months ago.
As an example I've been trying to get examples of a tiger face in a geometric, abstract style in black and white. I've done this before many times and Firefly used to be pretty good. Now its spewing out the same tiger head at different angles no matter what I add to the promted.It completely ignores the abstract and geometric. The vector and flat styles don't work at all like they used to. It still produces shaded and full color.

This is the prompt that use to work amazingly but its now just creating the attached. It was creating almost exactly this without the additional geomtric. I also added a style reference.

"tiger roaring logo, simple, angular, geometric, isolated on a white background, single color, abstract"

[droopydog500]

Hello @Offer-Readman,

Is this close to what you are looking for?

tiger roaring logo, simple, angular, geometric, isolated on a white background, single color, abstra... 

I find Model 2 works better with these kind of prompts than Model 3 does.

    droopy

[Offer-Readman]

Thanks @droopydog500 but not really. Thats simply adding a bit of an angular element to a basic tiger face. Its really gotten much worse and simply ignores large parts of the prompt. 

I used to use it for patterns and recently wanted tiger stripes but no matter how you word the prompt it give you a tiger head. It 100% didn;t used to be like this.

[Offer-Readman]

Heres an example prompt:
minimalistic and flat vector-style of tiger claw marks as if a tiger has scratched through a surface. bold, sharp, with clean, jagged edges. simple two-tone colour scheme, black scratches on a solid white background, modern, avoiding textures or gradients, clean, striking visual impact.
Heres the result using v3:

That pretty poor. v2 wouldn't do anything. Just kept spinning.

[Offer-Readman]

And another simpler prompt:
tiger stripes pattern for background, black and white, vector style, clean, no gradients, bold flat color

And heres the result using v3:

 This is basically the image you get in slightly different styles when you put the word tiger in the prompt. It just doesn't work any more.

[droopydog500]

Hi @Offer-Readman,

What do you mean by "v2 wouldn't do anything. Just kept spinning"?  I have not seen that with model 2 and have found model 2 better for the types of images you describe.

    droopy