Adobe Firefly Image 3 Model in Generative Fill (beta)
The April 2024 release of Photoshop (Beta) on the desktop 25.9 previews an early look at the new Adobe Firefly Image 3 Model. This under-development AI technology which powers Photoshop’s generative AI features, will create more realistic and higher-quality content.Everything created with Generative Fill, Generative Expand, Generate Similar, Reference Image, and Generate Background in this beta build is using the new image model.
Feedback request:
- Please let us know what you think of the images produced by the AI features. Do the results meet your expectations?
- Tell us how you will use these results in your projects.
Compare the results with those created using the same methods in the general Photoshop release.
Respond in this thread or provide input via the Beta Feedback Panel; click the beaker, next to the blue Share button, in the Photoshop Options Bar.

Known Issues:
- If using a mask with opacity to create the selection, Generative Fill does not work as expected.
- Groups of people are not generated as expected.
- Using the words “realistic” or “hyper-realistic” in the prompt results in the opposite.
We sincerely thank you for your time and genuinely appreciate your feedback to help us improve this technology.

