AI Inpainting: How to Edit Any Part of Any Image

You have the perfect photo — except for the trash can in the background, the stranger photobombing the shot, or the power line cutting across an otherwise pristine sky. AI inpainting lets you select the offending area, describe what should replace it (or simply tell the AI to remove it), and get a natural-looking result in seconds.

Inpainting is not new. Photoshop's Content-Aware Fill has existed for over a decade. What is new is the quality. AI inpainting models in 2026 understand scene context, lighting, perspective, and texture at a level that makes their edits virtually undetectable. The difference between a manual clone stamp job and modern AI inpainting is the difference between a rough sketch and a photograph.

What Is Inpainting?

Inpainting is the process of reconstructing lost or selected regions of an image. You define a mask — the area you want to change — and the AI fills it in based on the surrounding context and optionally a text prompt describing what you want.

There are two primary modes:

1. Removal inpainting: You mask an object and the AI replaces it with a plausible continuation of the background. The trash can disappears and the sidewalk continues naturally.
2. Replacement inpainting: You mask an area and provide a prompt describing what should appear there. A plain wall becomes a window. A cloudy sky becomes a sunset.

Both use the same underlying technology. The difference is whether you guide the generation with a text prompt or let the model decide based purely on context.

When to Use Inpainting

Inpainting is the right tool whenever you need to modify a specific region of an image while keeping everything else unchanged. Common use cases include:

• Removing unwanted objects: People, signs, wires, litter, watermarks, timestamps
• Fixing photo imperfections: Blemishes, lens flare, sensor dust spots, scratches on scanned photos
• Replacing elements: Swapping a cloudy sky for blue, changing a wall color, updating text on a sign
• Extending compositions: Filling in areas after straightening a tilted horizon or correcting lens distortion
• Cleaning up AI-generated images: Fixing hands, correcting artifacts, adjusting details that the generator got wrong

Step-by-Step Inpainting Tutorial

Step 1: Choose Your Source Image

Start with the highest quality version of your image available. Inpainting works on any resolution, but higher-resolution sources give the AI more context to work with and produce more detailed fills.

If your image is low resolution, consider upscaling it first before inpainting. This gives the model more pixels to analyze around the masked region.

Step 2: Create Your Mask

The mask defines what gets replaced. In Arteza's inpainting tool, you paint over the area you want to change using a brush tool. A few tips:

• Be generous with the mask: Include a small margin around the object you are removing. This gives the AI room to blend the new content naturally with the surrounding pixels.
• Follow object boundaries loosely: You do not need pixel-perfect precision. The AI is forgiving about mask edges.
• For removal, mask the entire object: Include shadows and reflections cast by the object. Leaving an orphaned shadow is the most common tell that an object was removed.
• For replacement, mask exactly the area you want new content: Keep the mask tight to the region you want to change.

Step 3: Write Your Prompt (Optional)

If you are simply removing an object, no prompt is needed. The AI will fill the area with a natural continuation of the background.

If you are replacing content, write a clear, specific prompt. Good prompts describe:

• What should appear in the masked area
• Style and material if relevant (wooden door, glass window, brick wall)
• Lighting conditions to match the rest of the image (sunlit, overcast, indoor warm lighting)

Good prompt example: "A large arched window with sunlight streaming through, matching the stone wall texture"

Weak prompt example: "window"

The more context you provide, the better the result will match the surrounding image.

Step 4: Generate and Review

Run the inpainting and inspect the result at full resolution. Pay attention to:

• Blending at mask edges: The transition between original and generated content should be invisible
• Lighting consistency: Generated content should match the light direction and color temperature of the original
• Perspective accuracy: New elements should follow the same vanishing points as existing ones
• Texture continuity: Surfaces that cross the mask boundary should look continuous

If the result is not perfect, regenerate. AI inpainting is probabilistic — each generation produces a slightly different result. Often the third or fourth attempt nails it.

Step 5: Fine-Tune if Needed

For complex edits, you may need multiple inpainting passes. Remove the main object first, then do a second pass to clean up any remaining artifacts. This layered approach often produces cleaner results than trying to fix everything in a single mask.

Advanced Inpainting Techniques

Iterative Refinement

For difficult removals — a person partially occluding a complex background — work in stages. Remove the person first, then refine the background in a second pass. Each pass lets the AI focus on a simpler problem.

Using Reference Areas

If you need the inpainted region to match a specific texture or pattern elsewhere in the image, make your mask overlap slightly with a good example of that texture. The AI will pick up on the pattern and extend it naturally.

Combining With Other Tools

Inpainting is most powerful as part of a workflow. A common pipeline:

1. Remove the background from a product shot
2. Inpaint any artifacts along the edges
3. Upscale the result to publication resolution
4. Outpaint to extend the canvas for a different aspect ratio

Each step builds on the previous one, and the AI handles each transformation seamlessly.

Common Inpainting Mistakes

Mask too small: If your mask barely covers the object, the AI often leaves visible traces at the edges. Extend the mask 10-20 pixels beyond the object boundary.

Forgetting shadows: Removing a person but leaving their shadow is an immediate giveaway. Always include cast shadows in your mask.

Inconsistent lighting in prompts: If you prompt for a "bright sunny window" in a room lit by warm tungsten light, the result will look pasted in. Match your prompt to the existing lighting.

Too many changes at once: Masking half the image and asking for wholesale changes usually produces inconsistent results. Work in smaller, focused regions.

The Quality Gap Has Closed

Five years ago, AI inpainting was a novelty — fun for demos but unusable for professional work. Today, the quality gap between AI inpainting and expert manual retouching has effectively closed for most use cases. A skilled Photoshop artist can still outperform AI on the most challenging edits, but for 95% of real-world inpainting tasks, AI delivers results that are indistinguishable from manual work in a fraction of the time.

The practical implication is straightforward: if you are spending more than 30 seconds on any object removal or content replacement in Photoshop, try AI inpainting first. You will often get a better result faster, and you can always fall back to manual tools for the rare cases that need them.