What Generative Fill Does and When to Use It

Generative Fill is a feature in Photoshop that uses artificial intelligence to remove unwanted objects from photos or add new elements to fill empty space. You select the area you want to change, and the tool generates new pixels that match the surrounding image — so you can remove a person from a crowd, erase a power line from a landscape, or extend a background to fill a larger canvas.

It works best on backgrounds and textures rather than complex subjects. A cluttered street scene, a distracting shadow, or an unwanted reflection are good candidates. A person's face or a detailed object in sharp focus is harder for the tool to handle convincingly, though results improve with each Photoshop update.

You need Photoshop 2024 or later and an active Creative Cloud subscription to access Generative Fill. The feature also requires an internet connection because the processing happens on Adobe's servers, not on your computer.

Key Takeaways

  • Generative Fill works by selecting an area of your image and letting the AI fill it with pixels that match the surrounding content.
  • The tool is most reliable for removing backgrounds, textures, and straightforward objects — less reliable for faces and fine details.
  • You can undo and regenerate results multiple times until you get an output you want to keep.
  • Photoshop 2024 or later with an active Creative Cloud subscription is required, along with an internet connection.
  • Generative Fill works alongside traditional selection tools like the lasso and magic wand, so you control exactly what area gets filled.

How to Select the Area You Want to Fill or Remove

Before you use Generative Fill, you need to tell Photoshop which part of the image to work on. Open your photo in Photoshop and choose a selection tool from the toolbar on the left. The lasso tool lets you draw freehand around the object you want to remove. The magic wand (called "Select Subject" in newer versions) automatically selects similar colors and textures. The rectangular marquee works for straight-edged areas.

Draw or click around the object you want to remove or the empty space you want to fill. You do not need to be perfectly precise — Generative Fill works better when your selection is slightly loose, including a little bit of the surrounding area. If you make a mistake, press Ctrl+Z (Windows) or Command+Z (Mac) to undo and try again.

Once you have selected the area, you can refine the edge. Go to Select menu and choose Modify, then Feather. A feathered edge (usually 5 to 10 pixels) makes the fill blend more smoothly into the rest of the image. Click Select and then Inverse if you want to select everything except the object — this is useful when you want to keep the object and remove the background around it.

Opening Generative Fill and Creating Your First Result

With your selection active, go to the Edit menu at the top and look for Generative Fill. Click it. A panel will open on the right side of your screen showing your selection and a text box that says "Describe what you want to generate." You can leave this blank and let the AI infer what should fill the space based on the surrounding pixels, or you can type a description like "blue sky" or "brick wall" to guide the result.

Click Generate. Photoshop will process your selection and show you the result. This usually takes 10 to 30 seconds depending on the size of your selection and how busy Adobe's servers are. The filled area will appear in your image, and you will see it as a new layer in your Layers panel on the right.

If the result does not look right, do not accept it yet. You can click Generate again to create a different version. Photoshop will show you up to three variations at once, and you can click on each one to preview it in your image. Keep generating until you see a result you want to keep.

Accepting, Rejecting, and Refining Your Fill

When you see a result you like, click the checkmark button to accept it. The filled area becomes permanent in your image, and the Generative Fill panel closes. If none of the generated versions work, click the X button to reject all of them and start over with a different selection or a new description.

Sometimes the fill looks good in the center but blends poorly at the edges. If that happens, undo the fill (Ctrl+Z or Command+Z), adjust your selection to include more of the surrounding area, and try again. A slightly larger selection often produces a more convincing blend.

You can also use Generative Fill multiple times on the same image. Remove one object, accept the result, make a new selection around a different object, and repeat. Each fill creates a new layer, so you can always undo individual steps if something does not work out.

Common Problems and How to Fix Them

If Generative Fill produces a blurry or distorted result, your selection was probably too large or included too much complex detail. Undo, make a smaller or more precise selection, and try again. If the fill does not match the surrounding colors or texture, try typing a description in the text box — "dark asphalt" or "green grass" — to guide the AI toward the right look.

If the tool seems slow or times out, check your internet connection. Generative Fill requires a stable connection to Adobe's servers. If you are on a slow connection, try closing other programs and tabs that might be using bandwidth, then try again.

Sometimes the AI fills your selection with something completely wrong — a face where there should be sky, or a texture that does not match at all. This usually means your selection included conflicting information. Undo, simplify your selection, and regenerate. You can also try a different description in the text box to steer the result in the right direction.

When Generative Fill Works Well and When It Does Not

Generative Fill is most reliable on uniform backgrounds like sky, water, grass, or plain walls. Removing a person from a blue sky background usually works on the first or second try. Removing a power line from a landscape often succeeds. Extending a brick wall or a wooden floor to fill empty canvas space usually produces convincing results.

The tool struggles with faces, hands, and fine details. If you try to remove a person's face or fill in a missing eye, the results are often unconvincing or distorted. Complex scenes with many overlapping objects are also harder — the AI can get confused about what should fill the space. In these cases, you may need to use traditional Photoshop tools like the Clone Stamp or Healing Brush instead.

Generative Fill also works better on photos than on illustrations or graphic designs. The AI was trained on photographs, so it understands natural lighting, texture, and color gradients better than it understands flat colors or stylized art.

Saving Your Work After Using Generative Fill

After you accept a Generative Fill result, your image contains a new layer with the filled content. Before you save, flatten your image if you want a single layer. Go to Image menu and choose Flatten Image. This merges all layers into one.

Save your work as a PSD file first (File > Save As, then choose Photoshop format). This preserves all your layers so you can edit further later. Then export a final version as a JPG or PNG for sharing or printing. Go to File > Export As, choose your format, and click Export.

Keep in mind that Generative Fill results are generated by AI and may not be perfect. Always review your final image carefully before sharing it, especially if it will be used for professional or commercial purposes. If the fill looks unconvincing on close inspection, undo and try a different approach.

Frequently Asked Questions

Do I need Photoshop 2024 or can I use an older version?

Generative Fill requires Photoshop 2024 or later. If you have an older version, you can update through the Creative Cloud app. Open Creative Cloud, find Photoshop in your apps list, and click Update.

What if I do not have an internet connection?

Generative Fill will not work without internet because the processing happens on Adobe's servers. You need a stable connection to generate results. If you are offline, you can use traditional Photoshop tools like Clone Stamp or Healing Brush instead.

Can I use Generative Fill to add people or objects to a photo?

Yes, but results are less reliable than removal. You can select an empty area and type a description like "person standing" or "dog running," and the AI will generate content. Results are often more convincing when you describe straightforward objects or backgrounds rather than complex subjects.

How many times can I regenerate a result before I have to accept it?

You can regenerate as many times as you want. Click Generate repeatedly to see different variations. There is no limit, though each generation takes processing time. Keep going until you see a result you want to keep.

Will Generative Fill work on a very small selection?

Small selections usually work better than large ones because there is less area for the AI to fill and more surrounding context to reference. If your selection is very small (a few pixels), the tool may struggle. If it is very large (most of your image), results are often less convincing.