How to Fix Blurry Pictures

Most "unblur" tools skip the diagnosis and go straight to the sharpening slider, which is why so many people end up with a photo that is crunchier than before and no clearer. Zoom in first. Thirty seconds of looking at the actual pixels will tell you which of the three problems you have, and that decides everything that follows.

Which Kind of Blur Do You Have?

Open the photo at full size and look at a hard edge, something like a window frame, a jawline or a line of text. What it looks like up close tells you what happened.

Why Does Sharpening Make It Worse?

The sharpen tool in every photo app does one thing: it increases contrast along edges. Light side gets lighter, dark side gets darker. On a slightly soft photo that reads as crispness. On a shaken photo it draws a bright halo around a smear, and on a compressed photo it turns the faint square patches into obvious ones. Neither has added a single pixel of real information.

There is a more clever traditional method called deconvolution, which tries to work out the path the camera took during the exposure and reverse it. It works on lab test charts and on a very small number of real photos. Adobe shipped a version of it in Photoshop as the Shake Reduction filter for nine years, then removed it in 2022, which tells you how often it helped.

What Does AI Actually Do to a Blurry Photo?

This is the part worth understanding before you upload anything, because it explains both why AI works so much better than sharpening and why it occasionally produces something unsettling.

An AI upscaler is not sharpening your photo. It is looking at your photo and drawing a new one that matches it. The model has seen millions of pictures of brickwork, hair, fabric, leaves and skin, so when it sees a blurred patch it knows what that patch probably looked like sharp, and it paints that in. For textures this is almost magic. A soft photo of a stone wall comes back looking like it was shot on a proper camera, because there is only one plausible way for a stone wall to look.

Faces are different, because there is only one correct way for a particular person to look and the model does not know it. On a mildly soft portrait the guess is close enough that nobody would notice. On a face that is tiny or heavily blurred, the model fills in eyes, teeth and skin texture that belong to a plausible person rather than the actual one. The result is sharp, and slightly wrong. The same applies to text, logos and number plates, where the model will produce confident letters that may not be the real ones.

What Can Be Rescued, Ranked

In rough order of how pleased you will be with the result.

• Small, compressed or screenshotted photos. The best case. There is real detail in there, just not enough of it, and the model has plenty to work from. Old Facebook downloads, WhatsApp forwards and images lifted from a web page all fall in here.
• Slight softness and mild missed focus. Very good results. Edges come back, skin texture returns, and faces stay themselves because the original still held most of the information.
• Low light noise. Good. The model removes the speckle and rebuilds the detail underneath it, usually better than the phone's own noise reduction managed.
• Mild camera shake. Usable. The smear is short enough that the model can find the edges. Check faces carefully.
• Heavy motion blur. Not really. The result will be sharp and clean and the person in it will be a stranger who looks a bit like them. Ask for a different photo.
• Tiny thumbnails under about 200 pixels. The model can make a nice picture, but it is a picture of someone it made up. Fine for a texture or a landscape, not for a person you know.

How to Fix a Blurry Picture With NeonSnap

What About Old Prints and Damaged Photos?

A scan of a print from the eighties has all the problems above at once, plus fading, colour casts, creases and the grain of the original film. Upscaling alone tends to sharpen the creases along with everything else. That is a restoration job rather than an upscale, and the two tools are separate for a reason.

What the Other Options Are

Honest comparison, since you may already have one of these to hand.

• Your phone. Google Photos on a Pixel has a Photo Unblur button that does a decent job on mild shake and softness. Samsung has a similar remaster option. iPhones do not have a true unblur, only sharpening.
• Lightroom and Photoshop. The Enhance menu has a Super Resolution option that doubles the pixel dimensions with reasonable intelligence. It is conservative, so faces stay faithful, but it will not rescue anything genuinely blurred. It also needs a subscription.
• Desktop AI apps. Topaz Photo AI and its competitors are the most controllable option, with separate sliders for sharpening, noise and upscaling, and they are the right choice if you have hundreds of photos to fix. They are paid software and take a while to learn.
• Browser tools like NeonSnap. One upload, one result, free to try. Less control, no install, and the right answer for the one photo you need fixed by tonight.

How to Stop It Happening Next Time

Nothing here is new, but it is worth a paragraph because every fix above is a compromise and the original is always better.

Tap the screen on the face before you shoot, so the phone focuses where you want rather than on the lamp behind them. Hold the phone with both hands and brace your elbows, and if it is dark, use night mode and stand still for the two seconds it asks for. Do not use the digital zoom on a phone, ever. It is a crop dressed up as a zoom, and it throws away exactly the pixels you will later wish you had. Walk closer instead. And when the photo is good, send the original file rather than the compressed copy your messaging app makes, so that the version everyone keeps is the sharp one.

If you are taking photos of yourself, our guide on how to look good in photos covers the lighting and angle side of the same problem, and the headshot guide goes into getting a properly sharp, usable portrait from a phone.