How to Use Generative AI in Photoshop

Photoshop's generative AI features have turned hours of manual compositing into 30-second operations. Need to extend a background, remove an object, or add elements that were never in the photo? Generative Fill and Generative Expand handle it with a text prompt and a click.

If you want to learn how to use generative AI in Photoshop effectively, the good news is that the technology has improved dramatically since Adobe shipped Generative Fill in mid-2023. As of early 2026, Photoshop includes multiple AI models (Adobe's own Firefly, plus partner models like Gemini 2.5 Flash Image and FLUX Kontext Pro), giving you different options depending on the output style you need.

This guide walks through every generative AI feature in Photoshop, with step-by-step workflows and the prompting techniques that produce the best results.

What You Need to Use Generative AI in Photoshop

Generative AI is available in Photoshop version 25.0 and later (the 2024 release onward). Here is what you need:

Photoshop subscription. Any Creative Cloud plan that includes Photoshop gives you access to generative AI features. The Photography plan starts at $9.99/month.

Generative credits. Each generation uses 1 credit. Your monthly allotment depends on your plan: the Photography plan includes credits, and Creative Cloud All Apps includes more. If you run out, credits refresh monthly, or you can purchase add-on packs (2,000, 4,000, 7,000, or 50,000 credits). Through March 16, 2026, Firefly Pro and Premium subscribers get unlimited generations.

Internet connection. Generative Fill processes through Adobe's cloud servers, so you need to be online when generating. Once the output appears, you can work offline.

Commercial rights. Content generated with Adobe Firefly models is cleared for commercial use. This applies to all non-beta generative features in Photoshop, so you can use the outputs in client work, marketing materials, and products.

If you want to build a solid foundation in AI-powered creative tools like Photoshop's generative features, the AI Academy covers the skills you need in a practical, hands-on format.

Generative Fill: Add, Remove, or Replace Anything

Generative Fill is the most versatile AI feature in Photoshop. You select an area, type a prompt (or leave it blank), and Photoshop generates content that blends with the surrounding image.

Step-by-Step: Removing an Object

1. Select the area. Use the Lasso Tool, Object Selection Tool, or Selection Brush to select the object you want to remove. Include a small margin around the object for cleaner blending.
2. Open Generative Fill. The Contextual Task Bar appears near your selection; click "Generative Fill." If you do not see the bar, go to Window > Contextual Task Bar.
3. Leave the prompt empty. For removals, an empty prompt tells Photoshop to fill the area based on surrounding context. This usually produces the cleanest result.
4. Click Generate. Photoshop produces three variations on a new generative layer. Click through them in the Properties panel to find the best match.
5. Refine if needed. If edges look off, select the problematic area and generate again. Each generation creates a separate layer, so you can mask or blend between them.

Prompting tip for removals: Empty prompts work best for removing objects from natural scenes. If Photoshop fills the area with something unwanted, try a short descriptive prompt like "grass and dirt path" or "clear blue sky" to guide the result.

Step-by-Step: Adding an Object

1. Select the target area. Draw a selection where you want the new element to appear. The size and shape of your selection influences the output; a tall narrow selection works better for adding a person than a wide square one.
2. Open Generative Fill from the Contextual Task Bar.
3. Enter your prompt. Be specific but concise: "golden retriever sitting" works better than "a very cute fluffy golden retriever dog sitting down happily on the grass looking at the camera."
4. Generate and review. Photoshop generates three options. Each appears on its own layer with a mask, so you can fine-tune placement and blending.

Prompting tip for additions: Include surface and lighting cues when they matter. "Person walking, afternoon sunlight from left" gives Photoshop context to match existing lighting. Avoid style instructions like "photorealistic" or "8K," since Photoshop already matches the photo's style.

Step-by-Step: Replacing a Background

1. Select the subject. Go to Select > Subject. Photoshop's AI detects and selects the main subject automatically.
2. Invert the selection. Go to Select > Inverse so the background is selected instead of the subject.
3. Open Generative Fill and enter your prompt. Describe the new background: "modern office with large windows" or "tropical beach at sunset."
4. Generate. The new background appears behind your subject with proper edge blending.

This workflow replaces what used to be 30-45 minutes of manual masking and compositing. If you are working on marketing visuals or social media content, combining this with the techniques in our guide to AI image creation gives you a powerful creative pipeline.

Learning to combine these tools effectively is what separates casual users from pros. Our AI Academy teaches exactly that -- building complete creative workflows with AI, not just isolated tricks.

Generative Expand: Extend Any Image

Generative Expand lets you make images larger by generating new content beyond the original edges. This is invaluable when you need a horizontal crop from a vertical shot, or when a composition is too tight.

Step-by-Step: Extending an Image

1. Select the Crop Tool from the Tools panel (shortcut: C).
2. Drag the crop handles outward beyond the current image boundary. You will see a transparent area where new content will be generated.
3. Select Generative Expand from the Contextual Task Bar that appears.
4. Enter a prompt or leave blank. A blank prompt tells Photoshop to continue the existing scene naturally. Use a prompt only if you want specific elements in the expanded area, like "mountains in the distance" or "more trees."
5. Click Generate. Photoshop fills the expanded area with content that matches the original image's style, lighting, and perspective.

Best use cases: Turning portrait images into landscape format for website banners, extending product photography backgrounds, creating wider establishing shots, and rescuing tight compositions where important elements were cropped.

Practical tip: Expand in stages rather than all at once. Extending 20-30% at a time produces more coherent results than trying to double an image's width in one generation. Each stage costs the same single credit.

Generative Upscale: Enlarge Without Losing Detail

New in Photoshop 2026, Generative Upscale goes beyond traditional upscaling by regenerating image detail using AI. Instead of just interpolating pixels, it creates new detail to enhance sharpness and clarity.

How to Use It

1. Go to Image > Image Size or use the Enhance Detail panel.
2. Select Generative Upscale as your method.
3. Choose your upscale factor, up to 4x the original resolution.
4. Click Generate.

This is particularly useful for older images, screenshots, or web-resolution photos that need to be printed or displayed at larger sizes. The AI infers and generates fine details like skin texture, fabric patterns, and text that traditional upscaling smears into mush.

Working with Partner AI Models

Photoshop 2026 introduced partner AI models alongside Adobe Firefly. You can now choose between different models depending on your needs:

• Adobe Firefly: the default, commercially safe model trained on licensed content.
• Gemini 2.5 Flash Image: Google's model, available for different creative approaches.
• FLUX Kontext Pro: offers alternative generation styles.

To switch models, look for the model selector in the Generative Fill options. Each model has different strengths, so experiment with the same prompt across models to see which produces the best result for your specific use case.

Prompting Tips for Generative AI in Photoshop

After extensive testing, these prompting strategies consistently produce better results:

Be specific about what, not how. "Red sports car parked on cobblestone street" outperforms "hyper-detailed ultra-realistic photograph of a sports car, 8K resolution, award-winning." Photoshop already matches the photo's look.

Reference the existing scene. If your photo has warm afternoon lighting, mention it: "woman reading a book, warm afternoon light." This helps the model maintain consistency.

Use simple language. Short, direct prompts (5-12 words) typically generate better results than long, elaborate descriptions. "Wooden bookshelf with books" is better than "an antique mahogany wooden bookshelf densely packed with colorful vintage hardcover books."

Iterate with variations. Each click of Generate produces three options. If none are perfect, refine your prompt slightly and generate again. The generative layer system means you never lose previous attempts.

Match perspective. When adding objects, consider the camera angle of your photo. "Coffee cup on table, viewed from above" for a top-down shot produces much more convincing results than a generic "coffee cup" prompt.

Limitations of Generative AI in Photoshop

Generative AI in Photoshop is powerful, but it has clear boundaries:

• Text generation is unreliable. The AI struggles with readable text, signs, and logos. Add text manually using Photoshop's type tools.
• Specific faces cannot be generated. You cannot prompt "a photo of [specific person]," as the model generates generic faces. This is a deliberate safety measure.
• Very large expansions lose coherence. Expanding an image by more than 50% in one generation often produces inconsistent results. Work in stages.
• Complex anatomical details like hands and feet can still look wrong, though each model update improves this.
• File size matters. Very large files (100MP+) may time out during generation. Work at reasonable resolutions and upscale afterward.

Practical Workflows for Common Tasks

Product photography cleanup: Select > Subject to isolate the product, invert, Generative Fill with "clean white background." Then use Generative Expand to create different aspect ratios for website, social, and marketplace listings. If you are posting these across platforms, our guide to using AI for Instagram covers optimization for each format.

Social media content creation: Start with one strong photo, use Generative Expand to create variations in different aspect ratios (1:1 for Instagram, 9:16 for Stories, 16:9 for YouTube), and use Generative Fill to add seasonal or thematic elements.

Client presentation mockups: Place a product shot in a scene, select the surrounding area, and generate a contextual environment. This replaces expensive location shoots for initial concept presentations.

What It Costs in Practice

For most users, the credits included with a standard Photoshop subscription cover regular use. Each generation uses 1 credit, and you get three variations per generation. A typical photo editing session might use 5-15 credits depending on complexity.

If you are doing high-volume commercial work (batch-processing product photos or creating large content libraries), you may need additional credits. The 2,000-credit add-on pack is usually sufficient for heavy users on a monthly basis.

The real cost savings are in time. Tasks that required 30-60 minutes of manual compositing now take 2-3 minutes with Generative Fill. For professional photographers and designers, this translates to handling more clients or spending more time on creative decisions rather than technical execution.

If you want to master AI-powered design workflows and stay current as these tools evolve, the AI Academy keeps its curriculum updated with the latest features and techniques.

FAQ

Want to integrate AI across your full creative and professional workflow? Start your free 14-day trial →