7 min read
Lesson 3 of 5
Photoshop Generative Fill & Expand
From the course AI Photography & Photo Editing
Photoshop's Generative Fill and Generative Expand are among the most impressive AI features in any creative tool. They use Adobe Firefly to add, remove, or extend image content based on text prompts.
Generative Fill: Adding & Replacing
Select an area of your image, then describe what you want:
- Use any selection tool (lasso, marquee, magic wand)
- Click "Generative Fill" in the contextual toolbar
- Type a text prompt (or leave empty for smart fill)
- Click "Generate" — Photoshop creates 3 variations
- Browse variations and pick the best one
Examples:
- Select an empty field → "A golden retriever playing fetch"
- Select a person's outfit → "Red evening dress"
- Select a bare wall → "Bookshelf filled with colorful books"
- Select an empty sky → Leave prompt empty for natural sky fill
Generative Expand: Extending Images
Extend your canvas beyond the original image boundaries:
- Use the Crop tool to expand the canvas
- Select the empty area
- Click "Generative Fill" (with or without a prompt)
- AI generates content that seamlessly continues your image
This is incredibly useful for:
- Converting portrait photos to landscape for banners
- Adding headroom above a tightly cropped portrait
- Extending backgrounds for design layouts
- Creating wider establishing shots from tight compositions
Object Removal
Select the object you want to remove, leave the prompt empty, and Generate:
- Photoshop fills the area with contextually appropriate content
- Works remarkably well for complex scenes
- Multiple generation attempts may be needed for perfect results
- Better than Content-Aware Fill for complex removals
Best Practices for Generative Fill
- Select generously — Include some surrounding area in your selection for better blending
- Be specific in prompts — "Tabby cat sitting on the grass" beats "cat"
- Generate multiple times — Each generation is different; pick the best
- Use layers — Generative Fill creates new layers, so you can easily undo or mask
- Match lighting — Mention lighting conditions: "in warm sunset light" or "in shadow"
- Check edges — Zoom in to verify seamless blending at selection edges
- Combine with manual editing — Use clone stamp and healing brush to fix small imperfections
Reference Image (Beta)
Photoshop now supports reference images in Generative Fill:
- Upload a reference image alongside your text prompt
- Generative Fill tries to match the style, color palette, and aesthetic
- Great for maintaining visual consistency across a series of edits
Limitations
- Results can be inconsistent — some generations are perfect, others are clearly AI
- Doesn't work well with very specific or complex requests
- Human faces in generative fill can look uncanny
- Each generation uses cloud credits (limited on lower-tier plans)
- Not suitable for forensic or documentary photography where authenticity matters
Key Takeaways
- •Generative Fill adds, replaces, or removes content based on text prompts with 3 variations per generation
- •Generative Expand extends images beyond their original boundaries seamlessly
- •Select generously and be specific in prompts for better results
- •Generative Fill creates non-destructive layers — easy to undo or refine
- •Combine AI generation with manual editing tools for professional results
0 of 5 lessons marked complete
Progress is kept in this browser only.