Blur Text in Image Online
Blur text in image online free, with no signup. Drop or paste your screenshot and drag over the line you need gone. It opens on a solid block rather than a blur, and that is deliberate: text is the one thing in a picture that people have genuinely reconstructed from a blur, because the character set is small and the font is usually known. Nothing is uploaded. PNG, JPG or WebP up to 25 MB.
Drop the screenshot with the text in it
Opens on a solid block rather than a blur, because blurred text is the easiest kind to read back. Or press Ctrl + V to paste one straight from your clipboard. PNG, JPG or WebP, up to 25 MB. Edited in your browser, never uploaded.
You Searched for Blur
Use a Block Instead
Every other page in this category will happily blur your API key and let you find out later. Text is the one subject where the recovery attacks are real, tooled and about five years old, so this page opens on the method that works and explains itself.
Same editor, different job
Each of these opens with the tool set up for its own subject rather than for text.
Tools that pair well with this one
All free, all in your browser, no signup.
How to hide text in an image
Three steps, no account, 25 MB per image.
Paste your screenshot
Press Ctrl+V anywhere on this page, or drag a PNG, JPG or WebP onto the box. Nothing uploads either way.
Drag over the line of text
The solid block is already armed. Cover the whole line, a little taller than the characters, so no descenders or capitals poke out of the top or bottom.
Check the rest of the frame, then download
Look for the second thing: a username in a tab title, an email in a sidebar, a notification that arrived mid-capture. Export flattened, full resolution, no watermark.
The uncomfortable part
Why blurred text is the easiest thing in a picture to recover
This is not a theoretical risk and it is not a scare story. It is a solved problem with public tooling, and text is the subject it works best against.
The attack is simple enough to explain in a sentence. A pixelation is a deterministic function: each block is the average of the pixels underneath it, so the same input always gives the same output. If an attacker knows roughly what they are looking for, an account number, a token in a known format, text in a familiar interface font, they can render candidate strings, apply the same pixelation, and compare block values until the numbers match. The search space is small because the character set is small.
Blur is a different transform with the same weakness. A Gaussian blur is a convolution, and convolutions can be partially inverted. How much comes back depends entirely on how hard the blur was relative to the size of the characters. A light blur over small text is recoverable to a meaningful degree. A heavy blur over a large region genuinely destroys enough information that it is not practical.
What makes text special is not that these attacks are stronger against it. It is that text has a finite alphabet, a predictable font, a known baseline and a known layout, so every one of those facts narrows the guess. A blurred face has none of that structure to exploit. A blurred sixteen-digit number has all of it.
So the ranking, worst to best: a mosaic is the weakest and it is the one everybody reaches for. A light blur is nearly as weak. A heavy blur is fine in practice for anything that is not worth real effort. A solid block on a flattened export has nothing underneath it to reconstruct, and that is the one this page arms for you.
The actual workflow
Covering a key, an email or an account number in a screenshot
Most of what gets hidden in an image is a string of characters in a screenshot, and most of it is hidden badly because of how the workflow is shaped.
The normal path is: capture something, save it to disk, find it again, upload it to a website, cover the thing, download it, then remember to delete the two copies now sitting in your Downloads folder. That is six steps, four of which exist for the tool's benefit, with the sensitive file on your disk the whole time. Press Ctrl+V here instead and the editor opens with the screenshot already in it.
- API keys and tokens: Cover the whole string including the prefix. A visible sk_live_ or ghp_ tells an attacker exactly which service to try and how long the rest is, which is more help than it sounds.
- Email addresses: Cover the local part and the domain. Leaving the domain visible in a support screenshot identifies the company, and in a small company that often identifies the person.
- Account and card numbers: Cover everything except the last four digits if a form asked you to leave those visible. That specific request is common enough to have its own guide, linked below.
- Names in a chat log: One block per name, and check the avatars and the tab title too. A screenshot of a conversation usually names the same person four times in four different places.
- The browser chrome: The tab strip, the bookmarks bar and the URL are where most accidental leaks live, because nobody was looking at them when they framed the capture.
The tool stays armed between regions, so this is one drag each. Every block is its own layer, and the layers panel lists them, which is the fastest way to confirm you covered five things and not four before the screenshot goes into a public ticket.
Related jobs on the same screenshot
If you insist
How hard a blur has to be if you are set on blurring
There are good reasons to want a blur here. It looks softer, it keeps the page looking like a page, and a solid black bar in the middle of a marketing screenshot is visually loud. So: how far do you push it.
The rule of thumb that survives contact with reality is that the region should stop reading as text. Not soft text, not text you can almost make out if you squint, but a smudge with no letter shapes in it at all. If you can see where one word ends and the next begins, the word shapes are still carrying information, and word shapes plus a known font is most of what a reconstruction needs.
The strength slider here scales its radius to the size of the region rather than using a fixed pixel value, which is what makes that rule usable. A fixed radius annihilates a 40px account number and barely softens a 900px heading, so the same slider position would mean something different on every region and you would learn to distrust it. Here, high means high on both.
The honest summary: use a heavy blur for something you would rather nobody read, like a colleague's name in a screenshot you are sharing internally. Use the solid block for anything that would cost you money or access if it came out. The block is already armed on this page, so the safe path is the one that takes no extra clicks.
Built for the screenshot in your clipboard.
The file you are about to censor is the one you decided was too sensitive to post. It should not have to go to a server first.
Solid block, already armed
The one style with nothing left underneath it is the one in your hand when the editor opens. Your first drag over a token covers it permanently rather than softly.
Paste straight from the clipboard
Press Ctrl+V or Cmd+V and the editor opens with your screenshot in it. The sensitive file never has to sit in your Downloads folder waiting to be uploaded somewhere.
One drag per line
The tool stays armed between regions, so a support ticket with an account ID, an email and a session token in it is three drags rather than three trips back to the rail.
Strength that means the same thing everywhere
If you do use a blur, the radius scales to the region rather than being fixed, so a 40px account number and a 900px heading are both unreadable at the same slider position.
Nothing is uploaded
Your browser opens the file and covers the text locally. No server sees the key you are hiding, which is the entire reason to be careful about which tool you paste it into.
No metadata in the export
The download is redrawn onto a canvas and re-encoded, so the timestamp, device and any GPS data in the original do not survive into the file you send.
Frequently asked questions
Whether blurred text can be read, what to use instead, and where this stops.
Sometimes, and more often than people expect. It depends on how hard the blur was compared to the size of the characters. A light blur over small text can be meaningfully recovered by deconvolution, especially when the attacker knows the font and roughly what kind of string they are looking for.
A heavy blur over a large region is a different story: too much information is genuinely gone for a reconstruction to be practical. The problem is that the difference between those two cases is a slider position most people never touch.
Yes, and it is not close. A solid fill does not look at the pixels underneath it, so on a flattened export there is nothing left to reconstruct, at any zoom, with any tool. A blur and a mosaic are both transformations of the original pixels, which means some amount of the original is still encoded in what you exported. That is why the solid block is what this page arms by default.
Paste the screenshot here with Ctrl+V, drag the solid block over the whole key including its prefix, and export. Cover the prefix too: a visible sk_live_ or ghp_ tells anyone reading which service the key belongs to and how many characters they are looking for. And rotate the key anyway if it was ever visible in something you sent, because a covered screenshot does not undo the moment it was on screen in a call.
A solid block does, because there is nothing left in the file to read. A light blur may not, since the same reconstruction that works on other short strings works on an address, and addresses are unusually guessable: the format is fixed, the domain is often visible from context, and the local part is frequently a name that appears elsewhere in the same screenshot.
No. There is no OCR here and no automatic detection of keys or account numbers, so you drag over each one yourself. That is a real limitation and worth stating plainly rather than implying otherwise. It is also the reason the tool stays armed between regions: if you are placing them by hand, placing the fifth one should not cost more than placing the first.
No. Your browser opens the file and covers the text locally. Nothing is sent to a server, nothing is stored, and nothing is sitting somewhere waiting to be deleted. That matters more on this page than almost anywhere else on the site, because the thing you are covering is frequently a live credential.
Drag one block over the whole area if the lines are adjacent, or one per line if there is content between them you want to keep. The tool stays armed either way, and each block is a separate layer you can move, resize or delete without touching the others.
Free, unlimited, no account, no watermark and no export cap. The work happens in your browser rather than on our servers, so there is no per-image cost for us to pass on to you.
Cover it before you paste it
Press Ctrl+V, drag a block over the key, and download. Nothing gets uploaded, and the block has nothing underneath it.