Feedback on Adobe Firefly Image Generation Quality

[Nue Leaf Creative]

Hi Adobe Team,

I wanted to take a moment to share some honest feedback regarding Adobe Firefly’s image generation tool. Unfortunately, my recent experience has been quite disappointing. The overall quality, realism, and accuracy of the generated images fall far short of what I’ve come to expect from Adobe’s usually excellent creative tools.

The results are often inconsistent, distorted, or poorly composed, even when prompts are clear and detailed. Compared to other platforms, Firefly’s current outputs feel unpolished and not suitable for professional or client-facing work.

I’m a long-time Adobe user and genuinely want Firefly to succeed — it has so much potential. However, at this stage, it’s difficult to justify using it in a professional workflow. I’d love to see improvements in image quality, realism, lighting control, and prompt interpretation in future updates.

Thank you for taking this feedback seriously — I know Adobe values user input, and I hope these concerns help guide future enhancements.

Kind regards

[droopydog500]

Hello @Nue Leaf Creative,

Thank you for your feedback. 

The models continue to change and improve. Those of us who have been using Firefly since Model 1 have seen significant improvement. Each model has its strengths and weaknesses. I still use Model 3 for certain images because it has areas of strength over Model 4. So, selecting the right model helps.

Also, using reference images, composition and style, can improve generations, if you have them available. Even sketching something to use as a composition reference can help.

Prompting also plays a role. We would be happy to look at some of your prompts and the results to see if we can help you change prompts to get a better result.

Finally, there still are things AI gets wrong. It is best to provide direct detailed feedback to the people who manage the models. You can do that by clicking thumbs down and providing feedback:

My best,

    droopy

[hc_8917]

The results of Firefly are utterly ridiculous. Especially if you want people in the picutures. The recent diversity push makes most images absolutely useless. If i want images for an event in Japan or a local event in Schwarzwald i don't need hispanics or black people. Especially since it decides to make all people the same ethnicity in on picture - often even giving two people the same face. Generating a new row of pictures without changing the prompt gives you only slightly tweaked images of the row above. Not sure what's going on here, but this is not how generative AI is supposed to work. They make you pay the same but at best 1/4 of a row can be used.