Firefly in Photoshop: Generative Fill & Expand
From the course Adobe Firefly: AI-Powered Creative Suite
Photoshop's integration with Firefly brings AI generation directly into professional photo editing workflows.
Generative Fill
Select an area in your image and describe what should appear there:
- Use the selection tools (Lasso, Marquee, Object Selection) to define the area
- Click "Generative Fill" in the contextual taskbar
- Enter a prompt describing what to generate (or leave blank to remove)
- Review variations — Photoshop generates three options
- Iterate — generate more variations or refine your prompt
Use Cases
- Remove unwanted objects by selecting and generating with an empty prompt
- Add elements: "a red vintage car parked on the street"
- Change backgrounds: Select sky, generate "dramatic sunset clouds"
- Extend outfits: Select clothing area, describe new attire
Generative Expand
Extend your canvas and fill the new areas with AI-generated content:
- Use the Crop tool and drag beyond the image boundaries
- Click "Generative Expand"
- Optionally add a prompt to guide the expansion
- Photoshop seamlessly extends the scene
This is invaluable for:
- Converting portrait images to landscape format
- Adding breathing room around subjects
- Creating wider compositions for different aspect ratios
Non-Destructive Workflow
Firefly generates content on separate layers, preserving your original:
- Each generation creates a new Generative Layer
- Toggle visibility to compare options
- Stack multiple generations for complex compositions
- Always keep the original layer intact
Tips for Best Results
- Make clean, precise selections for better generation quality
- Use descriptive prompts that match the image's existing style and lighting
- Generate multiple variations and combine the best elements
- Feather selections slightly for smoother blending
Named Example: The Washington Post & VML Agency
VML (formerly Wunderman Thompson) integrated Firefly into its campaign production pipeline for Ford, using Generative Expand to adapt a single hero image to 12 different banner ad dimensions without re-shooting. The Washington Post's visual team uses Generative Fill to remove distracting background elements from editorial photography faster than traditional masking.
Common Pitfall: Using Generative Expand on images with complex edges (trees, hair, crowds) produces blending artifacts at the expansion boundary. In these cases, expand in two smaller steps rather than one large jump, giving the AI more context for each step.
Best Practice: Always generate at least three Generative Fill variations before accepting one. The first variation is not always the best — Firefly explores different interpretations, and the second or third is often more contextually appropriate.
Key Takeaways
- •Generative Fill adds, removes, or replaces content within selections using text prompts
- •Generative Expand seamlessly extends images beyond their original boundaries
- •All AI generations are non-destructive, created on separate layers for safe editing
- •Clean selections and descriptive prompts that match existing style produce the best results
Progress is kept in this browser only.