How to do it, step by step
- Clean the lens, every time The single most common cause of soft phone photos is a fingerprint or pocket lint on the lens. Wipe it on something soft before you shoot. This costs two seconds and fixes more photos than any editing tool.
- Give the sensor more light Blur is nearly always a shutter speed problem, and shutter speed is set by available light. Move towards a window, turn a lamp on, step out of the shadow. More light means a faster shutter, and a faster shutter means a sharper frame.
- Lock focus on the face before the moment Tap the face on screen to focus, and press and hold to lock focus and exposure so the camera stops hunting when the subject moves. Missed focus on a face is the most common blur on a phone and this prevents it.
- Brace rather than hold Two hands on the phone, elbows tucked into your ribs, and a wall, doorframe or table taking some of the weight. Breathe out and shoot at the end of the breath. This alone removes most camera shake.
- Shoot a burst, not a frame Slide the shutter button to the left on an iPhone, or hold it on Android, and take ten frames instead of one. For a moving subject, a burst is the only reliable way to get a sharp one, and you delete the rest.
- Check at full zoom before you leave Open the photo, pinch in on the eyes, and look. Ten seconds of checking while you can still take another one is worth more than every unblurring tool on this page combined.
Four kinds of blur, only one of which is fixable
Diagnose before you reach for a tool, because the four look similar on a phone screen and behave completely differently.
Camera shake. The whole frame is smeared in one direction, usually with a visible streak on point lights. The subject was still and you were not. This is the one type where software can genuinely help, because the blur is uniform across the frame and roughly describable, so sharpening and deconvolution have something to work with. Mild shake improves noticeably. Severe shake does not.
Subject motion. The background is sharp and the person is not. Your shutter was too slow for how fast they were moving. Nothing recovers this. The information about where their face was during those milliseconds does not exist as a single sharp image, because during the exposure it genuinely was in several places.
Missed focus. Something else in the frame is sharp, usually the background or the shoulder rather than the eyes. The lens recorded a different plane. Unrecoverable in any meaningful sense, and the most common cause of "why is my photo blurry" on a phone, because the camera focused on the nearest high-contrast thing rather than on the face.
Soft everywhere, no streak. Usually a smeared lens, heavy noise reduction in low light, or digital zoom. Cleaning the lens fixes the first and costs nothing, and it is worth checking before anything else.
Why software cannot undo it
Blur is not a layer sitting on top of a sharp photo. It is what was recorded. Mathematically, a blurred image is the sharp scene convolved with a blur kernel, and recovering the sharp scene means running that backwards. Two problems make it impossible in practice.
First, the blur destroys high spatial frequencies, which is where fine detail lives. Motion blur attenuates them so severely that inverting the operation becomes numerically unstable: tiny amounts of noise in the file amplify into enormous errors. The detail is not hidden. It is gone.
Second, you usually do not know the blur kernel. With a handheld phone photo, the shake path is unknown, so the software has to estimate both the blur and the original image from one file. Researchers call this blind deconvolution and describe it as severely ill-posed: there are more unknowns than there are measurements, so many different sharp images would produce the same blur. Software cannot tell you which one was yours because the file does not say.
So what do AI unblur tools do instead? They do not invert anything. They take the blurry input and generate a sharp image that could plausibly have produced it, using patterns learned from millions of other photographs. The best-known demonstration of where this leads is a research face upscaler that, given a pixellated portrait, produced a crisp and entirely fictional face, and did so with a bias towards white features regardless of the input. It looked like a restoration. It was a guess.
What the tools do, and when to use them anyway
None of this makes the tools useless. It makes them a stylistic choice rather than a repair. Here is what each actually does.
| Tool | What it really does | Works on | Do not use for |
|---|---|---|---|
| Sharpening in your phone editor | Raises local contrast at edges | Mild softness, slight shake | Anything already noisy; it exaggerates grain too |
| Snapseed Structure | Boosts mid-scale detail without adding content | Flat, soft-looking photos | Faces at high strength, which go crunchy |
| Google Photos Photo Unblur | Trained sharpening, moderately conservative | Mildly soft photos and old scans | Heavy motion blur, which it cannot invent its way out of |
| Remini and similar one-tap enhancers | Generative redraw of the face and detail | Making a bad photo look shareable | Anything where the face has to be that person's face |
| Upscayl and other upscalers | Generative super-resolution | Enlarging an already sharp photo | Fixing blur; enlarging a blurry photo gives you a bigger blurry photo with invented texture |
| ChatGPT or Gemini | Re-renders the whole image from your upload | Turning a blurry photo into an illustration | Any use where the result should still be a photograph of that moment |
The order worth trying on a salvageable photo: crop first, then exposure and contrast, then a small amount of sharpening, and stop. If it still is not right, it is not going to be.
When invented detail matters
Whether hallucinated detail is a problem depends entirely on what the photo is for, and the line is clearer than people assume.
It does not matter for a photo of friends going to a story, a picture in a group chat, a photo dump where the whole point is a scrapbook feel, or a small print seen at arm's length. Nobody is auditing the pores. If a one-tap enhancer makes an otherwise dead photo postable, use it.
It matters in four situations. Documents and text: an enhancer will render a confidently wrong digit or letter, and a legible wrong number is far more dangerous than an illegible right one. Faces that must be identified: generative face restoration invents a face, and it invents one shaped by its training data rather than by your photo. Anything evidential, including insurance claims and disputes, where AI-enhanced footage has already been excluded from court proceedings on the grounds that it created false detail through opaque methods. Large prints and archival copies, where a viewer will inspect closely and where the file may outlive everyone who remembers what the original looked like.
The safe habit covers all of it: keep the original, treat the enhanced version as a separate derived image rather than a replacement, and say so if you pass it on.
What to do with a blurry photo you cannot reshoot
Sometimes the moment is gone and the photo is all there is. Five things genuinely help, none of which involve pretending it is sharp.
Crop in, not out. Softness is a function of how large the photo is displayed. A tight crop on a face at story size can read perfectly well when the full frame does not. This is the highest-value move and it is free.
Convert to black and white. Removing colour removes one of the cues the eye uses to judge sharpness, and it reframes the image as atmospheric rather than failed. A genuinely blurry black and white photo reads as a choice.
Add grain, do not remove it. Fine grain over a soft image gives the eye texture to hold onto and disguises the mush. This is why film photographs tolerate softness that digital ones do not.
Fix the light instead. Lifting shadows and correcting white balance improves perceived quality more than any sharpening slider, because much of what people read as blur is actually low contrast.
Use it as a supporting frame. In a set of photos, one soft, moving, alive frame among sharp ones is a benefit. On its own it is a mistake. Put it third in the dump, not first.
Where ShutterAI fits
Our own app, named so you can discount it: ShutterAI attacks this at capture rather than afterwards, because that is the only place blur is actually solvable. Best Face takes a burst and picks the sharpest frame with everyone's eyes open, which is the single most effective anti-blur tool there is. Light Finder tells you when the light is good enough that your shutter will be fast. Vault keeps the original of everything.
It has not shipped. ShutterAI lands on iPhone this autumn, free at the outset, and the waiting list sits on the homepage. The capture steps below need no app at all.
Questions people ask
How to fix blurry photos?
Diagnose the blur first. Mild camera shake responds to cropping, contrast and light sharpening. Subject motion and missed focus are not recoverable, because the detail was never recorded. AI unblur tools generate plausible detail rather than restoring the original.
Do AI unblur apps actually work?
They produce a sharper-looking image, but by drawing new detail rather than recovering lost detail. That is acceptable for a social post and unsuitable for documents, identification, evidence or archival copies.
Why are my iPhone photos blurry?
In order of likelihood: a dirty lens, focus landing on the background instead of the face, too little light forcing a slow shutter, or the subject moving. The first two take seconds to rule out.
Can Google Photos fix a blurry photo?
Photo Unblur, free to all users, genuinely improves mildly soft photos and old scans. It cannot recover heavy motion blur, and no tool can, because the information is not in the file.
What is the best way to avoid blurry photos?
More light, locked focus, braced arms and a burst instead of a single frame. Those four habits prevent almost every blurry photo, and they are free.