AI Inpainting: How to Fix Torn and Missing Parts of Photos

What Inpainting Actually Means

The word "inpainting" sounds more technical than it really is. It just means filling in missing or damaged parts of an image. Traditionally, this was done by hand. An artist or retoucher would literally paint in the missing area, matching colors and textures by eye. Really skilled people could do amazing work, but it took hours and cost a fortune.

AI inpainting does the same thing, just automatically. The AI looks at the pixels surrounding the damaged area, analyzes patterns, textures, and colors, and then generates new pixels to fill in the gap. It's not copying and pasting from nearby areas (that's old-school clone stamping). It's actually creating new content that it thinks should logically be there based on context. And honestly, the results in 2026 are kind of wild compared to even two or three years ago.

How Modern AI Inpainting Works (Without the PhD)

I'll spare you the deep technical dive, but the basics are actually pretty intuitive. Modern inpainting uses something called diffusion models. The way to think about it: the AI has been trained on millions and millions of images, so it has a really strong understanding of what the world looks like. Grass looks like this. Sky gradients look like that. Brick walls have this kind of pattern. Shirt fabric folds in certain ways.

When you give it a photo with a missing section, it doesn't just look at the immediately surrounding pixels. It considers the entire image. It understands that if there's a garden scene with a fence and grass, the torn-off corner probably had more grass and maybe more fence. If someone's wearing a striped shirt and a chunk is missing from the middle, it knows to continue the stripe pattern. The model generates this fill iteratively, refining it step by step until the result blends seamlessly with the rest of the photo.

Some tools also use what's called generative fill, where the AI doesn't just extrapolate from context but can generate entirely new content. Adobe made this mainstream with Photoshop's Generative Fill feature. You select an area, and the AI creates plausible content to put there. It's powerful stuff.

What Works Well (and What Still Doesn't)

Here's the deal. AI inpainting is incredibly good at certain things and still noticeably limited at others. Knowing the difference will save you a lot of frustration.

Backgrounds and textures? Absolutely nailed. If the torn part of your photo was sky, grass, a wall, a table, carpet, pavement, pretty much any repeating or smooth pattern, the AI will fill it in so well you genuinely can't tell anything was ever missing. Same goes for clothing, foliage, and simple objects. The technology is remarkably good at continuing patterns and generating natural-looking textures.

Faces are a different story. If someone's face has a scratch across it or minor damage, AI inpainting can handle that beautifully. But if a big chunk of someone's face is completely gone, like the entire left side is torn away, the AI is essentially guessing what that person looked like. And it might create a perfectly realistic-looking face that just... isn't the right person. For family photos where identity matters, this is the big limitation. The AI can make it look good, but "good" and "accurate" aren't always the same thing.

Hands, text, and fine details like jewelry can also be tricky. But for the vast majority of torn-corner and ripped-edge scenarios? The results are genuinely impressive.

Best Tools for Photo Inpainting Right Now

If you want to do inpainting yourself, there are some solid options. Photoshop's Generative Fill is probably the most powerful if you're willing to pay for a Creative Cloud subscription. You select the damaged area, hit generate, and it gives you multiple options to choose from. The results are consistently good and you get a lot of control over the process.

For something free and browser-based, cleanup.pictures is surprisingly capable. You just paint over the area you want removed or filled in and it handles the rest. It's more aimed at removing unwanted objects than restoring torn photos, but it can work for simpler inpainting jobs. There's also Lama Cleaner (now called IOPaint), which is an open-source tool that runs locally and uses some of the same underlying models as the commercial tools.

For phone-based options, several apps have built inpainting into their restoration pipeline. You don't even have to manually select the damaged area because the app detects it and fills it automatically. That's obviously easier for people who don't want to mess around in Photoshop.

Walking Through a Real Example

Let me describe a scenario that's really common. You've got a family photo from the 1970s. It's a nice shot of your parents standing in front of their first house. But the bottom-right corner is torn off, and there's a crease running diagonally across the lower third. The torn area takes out part of the front yard and a section of the sidewalk.

First step is always scanning or photographing the damaged print at a decent resolution. Don't try to flatten or tape the torn pieces back together first, just scan it as-is. The AI works better when it can clearly see where the damage is rather than trying to work around a bad tape job.

When you run this through an inpainting tool, the AI identifies the torn edge and the missing area beyond it. It looks at what's still visible: grass, a concrete sidewalk, the edge of the yard. Then it generates a continuation of those elements into the missing space. The grass texture matches. The sidewalk continues at the right angle. The overall lighting stays consistent. In most cases, you end up with a photo that looks complete. Not perfect if you zoom in to 400%, maybe, but at normal viewing size it looks like the photo was never torn.

The crease is even simpler. The AI identifies the line of damage, looks at the pixels on either side, and fills in the crease with what should be there. For something like a crease over grass or a house wall, it's basically invisible after processing.

ClearPastAI Handles the Restoration, You Handle the Memories

Look, not everyone wants to learn Photoshop or spend time masking out damaged areas by hand. And that's totally fine. ClearPastAI was built specifically so you don't have to think about any of this technical stuff. You open the app, pick your damaged photo, and the AI handles everything: the tears, the creases, the missing corners, the scratches, all of it in one pass.

The whole point is that you shouldn't need to be a Photoshop expert to save a photo that means something to you. These are memories of people and moments that matter. The technology should just work so you can focus on the part that actually counts, which is having that photo back in a condition you can actually enjoy looking at and sharing with your family.