Skip to main content
MeredithStotzner
Community Manager
Community Manager
March 18, 2026

Firefly Image 5 (Preview) in Photoshop Generative Fill

  • March 18, 2026
  • 1 reply
  • 183 views

Firefly Image 5 enables a new instruction‑based editing workflow in Generative Fill and it's currently in Photoshop (Beta) for desktop
Firefly Image 5 is Adobe's most advanced image generation and editing model yet; you can edit images more intuitively and precisely with just a prompt. Simply describe your edits with natural language instructions to transform the entire image—such as changing time of day or modifying a subject’s pose - making the precise changes you want with high quality results, while keeping the rest of your pixels and the details of your original image the same.
 

How to try it:

1. Open an image and select the entire canvas (Select > Select All…).

(While smaller selections are supported, this beta performs best when the full image is selected.)

2. Click Generative Fill in the Contextual Task Bar.

3.Choose Firefly Image 5 (Preview) from the model picker in the Contextual Task Bar.

4. Enter a clear, instructional prompt (for example: “change the time of day to sunset”).

5. Click Generate.
 

The Contextual Task Bar. The model picker is set to Firefly Image 5 (Preview)
The model picker menu. Set to Firefly Image 5 (Preview)

Result:

One variation is returned in the properties panel.

Note: The way to write prompts for Firefly Image 5 is very different from other Firefly models. Here are a few tips:

  • Precise prompting gives better results. Use exact color, names, detailed descriptions, and clear action verbs instead of vague terms.
  • Explicitly prompt what should remain unchanged to protect important elements.
  • Quote the exact text you want to change: For example, Replace 'joy' with 'happy'.
  • Specify camera angle, position, or framing to prevent unwanted repositioning.
  • Choose specific, powerful, and clear action verbs.

Please visit the Beta Feedback panel to share your thoughts or rate the result in the Properties panel using the rating options. 

Thank you.

    1 reply

    robertw44482386
    Inspiring
    March 19, 2026

    This sounds so exciting. What are the resolution constraints? Would there be a benefit to creating smart objects of small sections of an image, opening that object as a full document and then performing the operation on that?