ReviewsFeaturesBeyond CollageShowcaseToolkitFAQBlogMake Amazing Collage Now
← Back to Tools
Image Tools

Pixelate Image Online

Hide faces, number plates and names under a mosaic, a blur or a solid box. Or turn a whole photo into pixel art. Auto find faces works on your device. Up to 50 photos go at once, free, with nothing uploaded.

FreeNo Sign-upNo WatermarkFinds Faces
Drop images here or click to upload
JPG · PNG · WEBP · GIF · AVIF · paste with Ctrl+V
What to Hide
Drag on the photo where you want to hide
Effect
Block shape
Block Size
Blocks
Measure in
Pixel Art Colours
Save As
🔒 Images never leave your device
Add an image first to preview it here

Make stunning posts from normal photos

Download CollagePhotoApp for free

GET IT ONGoogle PlayDownload on theApp Store

Quick answer: to pixelate an image, open it in the editor above. It starts on Marked areas, which hides only what you mark: drag a Box over a number plate or tap Auto find faces to cover every face. Whole photo turns the entire picture into blocks. Choose Pixelate, push the Block Size slider to the right until nothing reads and press Download. Everything is processed here itself, nothing is uploaded to our servers.

People who search for this want one of two things. Some need to hide something: a stranger's face, a child at the school gate, a name in a screenshot. Others want the look itself, a photo rebuilt from chunky squares like a screen from an old console. The editor does both. This page explains what works for each, with the numbers worked out by the same code the tool runs.

Start with the defaults. A marked face gets 6 blocks across its own width, whatever its size. Whole photo cuts a 4,032 × 3,024 phone photo, about 12.2 megapixels, into 64 blocks across. That makes 64 × 48 blocks of 63 pixels each. The rule for faces matters more than anything else here. The section on block size shows why.


How to Pixelate an Image

Five steps. About a minute for one photo, a little longer when you check every face in a group shot.

  1. Open your photos. Drop images onto the box above, click it to browse your device or press Ctrl+V to paste one. JPG, PNG, WebP, GIF, BMP and AVIF open in current browsers. HEIC opens in Safari and in every browser on an iPhone or iPad. A batch holds up to 50 images.
  2. Choose what to hide. Marked areas comes switched on under What to Hide, with All except marked and Whole photo beside it. Tap Auto find faces at the top of the controls to put an oval over every face it finds. To mark by hand, choose Box, Oval or Brush and drag on the photo.
  3. Pick an effect. Under Effect, choose Pixelate, Blur, Solid colour or Emoji. Pixelate also offers a Block shape of Square, Circle, Hexagon or Diamond.
  4. Set the strength. Move the Block Size slider to the right for bigger blocks and a stronger hide. The readout shows how many blocks run across and how big each one is in pixels. Press Original vs Pixelated and drag its line across the photo to check one against the other. Both appear again below at thumbnail size.
  5. Download. Pick a format under Save As and press Download. A batch gets Download N Images as ZIP instead. Each file keeps its own name with -pixelate-image-collagephotoapp.com added.

Nothing saves until you press Download. Drag marks around, try all four effects, slide the blocks from fine to coarse and back. Original vs Pixelated splits the preview with a line you drag: the untouched original on one side, the result on the other. Undo and Redo step back and forth through your marks. Clear all marks wipes them in one go.

Which option under What to Hide? Marked areas, the one the editor opens on, suits the everyday job of hiding two faces and a plate in an otherwise ordinary picture. All except marked is the reverse: the people you mark stay sharp and everything around them turns to blocks. Whole photo suits pixel art and the rare photo where nothing at all should be recognisable.

Two shortcuts save time. Paste a screenshot straight in with Ctrl+V, with no file to find first. When the job is done, Copy Image puts the result on your clipboard where the browser allows it, ready to paste into a chat or a document.


Hide a Face, a Plate or a Name

Hiding is the job most people arrive with. A street photo with strangers in it. A car on a driveway with its number plate in plain view. A screenshot of a group chat where only one message should show. Google Street View blurs faces and number plates across its imagery for the same reason: the picture is fine to share, the detail is not.

Marked areas, the option the editor opens on, puts the effect only where you mark. Each mark is a shape on the photo that you can select, move, resize, copy and delete. Marks belong to each photo, so a batch of 50 never mixes one photo's faces up with another's.

Find Faces on Your Device

Tap Auto find faces at the top of the controls and the tool looks for every face in the photo. It places an Oval mark over each one it finds.

The search runs on your device. This page reads the photo and gets the boxes back without anything leaving. Your photo travels nowhere.

Small faces are the hard part. Face finding works on a shrunken copy of the photo, so in a 4,032 pixel wide group shot a face in the back row ends up a handful of pixels across. The finder works around that in passes. It scans the whole photo first, then overlapping tiles, then smaller tiles again when the faces it finds are small. Each pass sees the back row from a little closer.

A detected box runs from brow to chin. Hair and ears give people away just as surely, so the oval the tool places is 40% wider and 65% taller than the box. It also sits a little higher, to take in the hairline.

We tested it on six public-domain group photos holding 67 faces, from 2 to 21 per photo. The set included side profiles and children half hidden behind adults. It boxed all 67. It also drew 3 boxes where there was no face, which you remove with × or the Delete key.

That is our own measurement on six photos. A small sample, not a promise about yours. The finder misses more on side profiles, faces turned away from the camera, faces in heavy shadow and very small faces. Treat it as a fast first pass. Then work across the photo from edge to edge and draw an Oval over anything it missed. Any face it skips goes out with your photo.

In a batch, Auto find faces in all photos runs the same scan on every image in turn. Step through the results afterwards. Each photo deserves its own look before it goes out.

Box, Oval and Brush

Each marking tool has an obvious job. Box suits anything with straight edges: number plates, name badges, the address label on a parcel, a line of text in a screenshot. Oval suits heads, which are rounder than any box and look less clumsy under a curved edge. Brush is for everything that refuses to be a rectangle: a tattoo that wraps round a forearm, a name printed around a mug, a house number on a gate photographed at an angle.

Drag on the photo to draw a Box or an Oval. The shape grows from the point you press and follows your finger or mouse, so a plain tap draws nothing. With Brush, paint over the area. A circle on the photo shows how wide the stroke will be before you start. The Brush size slider runs from 1% to 15% of the photo's short side, so one setting covers the same share of a small screenshot and a 12.2 megapixel photo. The box beside it takes an exact width in pixels.

Editing works the way you would hope. Tap a mark to select it and drag it to move it. Drag a corner or an edge to resize it. Press × or the Delete key to remove it. The copy button beside × adds a second mark of the same size and shape, handy for a row of number plates in a car park. The arrow keys nudge a selected mark, which helps when a box sits a few pixels off a plate and the last letter peeks out from under it.

Mark generously. A box drawn tight to a plate keeps the plate's exact outline and position on show. Draw it a little bigger than the thing it hides and the outline goes too. The same applies to faces: an oval that stops at the jaw leaves the chin and neck for anyone to read.

Plates have one more trap. A plate is high-contrast lettering on a plain background, so a plate pixelated with too many blocks keeps a readable rhythm of dark and light. Give a plate fewer blocks than you think it needs.

Selling something online brings the same jobs in a different order. A car photo for a listing needs the plate gone. A photo of a flat for rent needs the door number and the family photos on the shelf gone. A parcel you are reselling needs the old label gone. One Box each takes a few seconds. The listing then shows the thing for sale instead of the person selling it.

Screenshots carry more than the message you meant to share. A chat screenshot shows the other person's name and profile photo at the top. Sometimes it shows their phone number too. A notification can drop down across the top of the screen at the wrong moment, with a preview of somebody else's message in it. Mark each of those, not only the line you were worried about.

Keep the Subject Sharp

Sometimes the subject is the one person who should be seen, like a bride with forty guests behind her or a portrait shot on a busy street. A product on a cluttered desk has the same problem. All except marked flips the logic. Mark the subject and the tool hides everything else.

Here Block Size counts across the photo's whole width, not the mark. At the default 64, each block on a 4,032 pixel wide photo is 63 pixels. A stranger's face 120 pixels wide in the background gets 1.9 blocks across. Gone. A passer-by close to the camera, with a face 600 pixels wide, gets 9.5. That is more detail than a marked face gets at the default 6. Slide to 32 blocks across and each block grows to 126 pixels. The close stranger drops to 4.8.

Auto find faces helps here too. Run it, delete the ovals over strangers and keep the ones over your own people. All except marked then keeps your people sharp and hides everyone else. In a screenshot, mark the one message that matters and the rest of the chat turns to blocks around it.


How Big Should the Blocks Be?

Big enough that nobody can tell who it is. That sounds obvious until you notice how most tools measure a block: in pixels. A block size in pixels means nothing without the size of the face underneath it.

Take a block of 16 pixels. On a face 60 pixels wide in the back row, it leaves 3.8 blocks across, a smudge. On a close-up face 1,920 pixels wide, nearly half the width of a phone photo, the same block leaves 120 blocks across. Anyone who knows that person will see a slightly rough photo of them.

So the tool counts blocks across the mark instead. Under Block Size, Measure in has two settings: Blocks across and Pixels. Blocks across, the default, splits each mark's width into the number you choose. A small plate and a big face get the same strength of hide. A 240 pixel face at the default reads "6 blocks across each mark · 40 px". The slider runs from 2 to 64 blocks across a mark. Pixels works the old way, from 2 to 1024 pixels a block, for the times you need one exact size everywhere.

Pages that give a number for faces mostly settle on 6 to 8 blocks across. Our default is 6, the coarse end of that range. The table shows what those counts mean in pixels for faces of different sizes in a 4,032 × 3,024 photo. The last column shows what a fixed 16 pixel block would leave instead.

Face widthShare of the photo's widthBlock at 6 acrossBlock at 8 acrossBlocks across with 16 px blocks
60 px1.5%10 px7.5 px3.8
120 px3%20 px15 px7.5
240 px6%40 px30 px15
480 px11.9%80 px60 px30
960 px23.8%160 px120 px60
1,920 px47.6%320 px240 px120

Group photos are where this pays off. Faces in one shot can differ in size by ten times. A child at the front with a face 900 pixels wide and an uncle at the back at 90 pixels both get 6 blocks across, which works out at 150 pixel blocks for the child and 15 pixel blocks for the uncle. One setting hides both to the same degree. With one fixed pixel size, a block that looks neat on the uncle leaves the child 60 blocks across, plainly recognisable.

A long Brush stroke follows the same rule. Blocks across counts across the stroke's whole span, so a sweeping stroke along an arm gets big blocks and a small dab gets small ones. The blocks never slide while you paint. As the stroke grows they double in size, four merging into one right where they sit.

Pixels still earns its place. Say you are hiding the same field in thirty screenshots from one app, all the same size. A fixed block of 24 pixels gives every one of them the identical look, which reads as tidy on a help page or in a bug report. Faces are the case where it goes wrong, because faces never come in one size.

The research says the number matters less than people hope. In 2016 McPherson, Shokri and Shmatikov published Defeating Image Obfuscation with Deep Learning. They trained a neural network to recognise people in obscured photos. With blocks of 16 × 16 pixels, it identified faces 57% of the time among 530 people. It had the right person in its top five guesses 72% of the time. On a 40-person set of faces it scored above 95% for every mosaic size they tested, where random guessing scores 2.5%. It also beat YouTube's face blur.

People do well too. Lander, Bruce and Hill showed in 2001 that famous faces in TV clips stayed recognisable through heavy pixelation. Moving faces were easier to recognise than still ones.

No block count makes a face safe from someone who already knows it. A machine trained on the person can work with very little. So can a friend. Fewer blocks always beat more. For a stranger in a crowd, 6 is a sensible start. For someone your audience knows, such as a colleague, a neighbour or a child in their own school's photo, go down to 4 or 3. When the face must be gone for good, use Solid colour.

The Squint Test

A pixelated face can get easier to recognise when you make it smaller or blurrier. That sounds backwards until you see the experiment behind it.

In 1973 Leon Harmon and Bela Julesz published a paper in Science built on a block portrait of Abraham Lincoln, which Harmon took from a $5 bill. Coarse blocks made the portrait hard to recognise. Blurring the blocks made it easier. The sharp edges between blocks cause the trouble. They are loud, fine detail. They mask the soft, coarse shapes a face is really made of. Take the edges away and the face comes back. In 1976 Salvador Dalí based two works on Harmon's Lincoln.

You do this every day without trying. A feed shows your photo as a small thumbnail. A friend glances at your screen from across the table. Both shrink the blocks until their edges disappear. That is exactly when a face drops out of hiding.

Original vs Pixelated, in the preview header, runs this check for you. Drag its line across a face to see both versions on the same spot. Below the preview, both appear again at thumbnail size. In Marked areas they zoom to the part you marked. If you can name the person in the small Pixelated copy, someone who knows them can too. Push the slider to the right and look again.


Pixelate, Blur, Solid Colour or Emoji

Effect decides what goes inside a mark. All four hide detail. Only one removes it completely.

EffectWhat it doesWhat still showsUse it for
PixelateSplits the area into blocks and paints each block its average colour. Blocks can be Square, Circle, Hexagon or DiamondColours, rough shapes and the outline of the area. Familiar faces and moving footage leak the mostStrangers in a crowd, faces in social posts, the classic censored look
BlurSoftens the area. Blur Strength takes the place of Block SizeOverall shapes and colours. The 2016 study above also identified people through YouTube's face blurBusy backgrounds and a gentle look, where tidiness matters more than secrecy
Solid colourFills the area with Black, White, Average or a colour you choose. Average uses each mark's own average colourThe size and position of the box. Nothing inside itPasswords, card numbers, account numbers, addresses, anything typed
EmojiCovers each mark with one of 16 emoji or Any emoji you type, over a heavy pixelateHeavily pixelated scraps around the emoji's edgeChildren at a party, family posts, anything that should feel friendly. Marks only

Pixelate is the effect everyone recognises. Viewers read blocks as "hidden on purpose", which helps in itself: nobody mistakes a censored face for a bad photo. Wikipedia's entry on pixelization names its weak spot. In moving pictures, someone can combine the changing blocks from frame after frame to rebuild a sharper image. The same entry points out that a solid fill avoids the problem.

Blur reads as softer and calmer. It suits a crowd or a messy background when the aim is a tidy picture more than a secret. For a face that must not be recognised, pick Pixelate with few blocks or Solid colour instead.

Solid colour is the honest one. A black box removes everything inside it, so nothing is left to average, sharpen or rebuild. Average makes it less loud: the box takes the mark's own average colour and sits in the photo like a patch of wall. White suits a scanned letter. A colour of your own can match a brand or a background.

Emoji is the friendly one. A sunglasses face over a toddler at a birthday party says "private" without the crime-scene feel of a black bar. Emoji works on marks only, never on Whole photo. Under every emoji the tool lays a heavy pixelate, because an emoji is not a square. Its transparent corners would otherwise let bits of the face show through.

The 16 built-in choices are 😎 🙂 😂 😇 🤡 🥸 🐱 🐶 🐸 ⭐ ❤️ 🙈 🌸 👾 🔒 ⬛. Any emoji takes whatever your keyboard can type, such as a football over a football club's crest. One emoji covers every mark on a photo, which keeps a group shot looking deliberate rather than random.

Why Pixelated Text Is the Weak Spot

Faces vary endlessly. Text does not. A screen draws every letter "a" in the same font, at the same size, on the same pixel grid. That sameness is what breaks pixelated text.

In December 2020 Sipke Mellema released Depix, a tool that recovers text pixelated with a linear box filter. The attacker types out a De Bruijn sequence, a string that holds every short combination of characters, in the same editor and font as the target. Depix pixelates a screenshot of that string the same way and matches it block against block with the hidden text until the words fall out. It works on screenshots, not on photographs. Its code is public on GitHub for anyone to run.

Short secrets fare worse still. A four-digit PIN has 10,000 possible values. Someone can type out every one in the same font, pixelate each at the same block size and compare. It takes patience rather than clever software.

Our blocks average in linear light, which changes the numbers inside each block. Don't treat that as protection. Anyone who knows the method can copy it.

For passwords, card numbers, account numbers, addresses and names in a screenshot, use Solid colour. Draw the box a little wider than the text as well. A bar cut exactly to a word still tells everyone how long the word is.


What Pixelation Does Not Hide

A face is one clue among many. In 2007 Demanet and four colleagues found that masking a familiar person's face did not fully stop recognition, because viewers also used hair, clothing and other cues. Anyone who knows the person has the same shortcuts.

Before you post, look at the photo the way a nosy neighbour would. The giveaways pile up quickly:

  • Hair and build. A bright red ponytail or a very tall frame identifies someone from across a car park.
  • Clothes and accessories. A favourite jacket, a distinctive bag, a cast on an arm.
  • Tattoos, scars and birthmarks. Mark them with Brush. They follow the person into every photo.
  • Badges and uniforms. A school crest, a club shirt, a lanyard with a printed name, a company logo on a polo shirt.
  • Reflections. Sunglasses, a window behind the subject, a mirror, the glossy door of a car. Faces and screens show up in all of them.
  • The scene. A house number, a street sign, the view from a window. The face is hidden. The address is not.
  • Words around the photo. A caption, a tag or a comment can name the person you just hid.
  • Voices. A still from a video carries no sound, but the clip it came from does. A voice can give a person away as fast as a face.

Picture a parent posting the under-nines football final. Every child's face sits under a neat oval. The team name runs across every shirt. The club crest hangs on the banner behind the goal and the date sits in the caption. Anyone local now knows which children, which club and which Saturday. Pixelate the crest and the banner too. Or crop them out first with the Image Aspect Ratio Changer.

Moving pictures leak more again. Lander, Bruce and Hill found moving faces easier to recognise than still ones. In an animated GIF a mark stays in one place on every frame. It does not follow a face that moves. Draw the mark over the whole path the face travels or use Whole photo.

Location Data and the Original File

The pixelated file is half the story. The other half is the original.

The tool encodes every file it saves fresh from pixels. The file carries none of the original's metadata: no camera model, no date stamp, no GPS position. A photo taken in your front garden, pixelated and saved here, no longer says where the garden is.

The original stays untouched. That is the catch. It still holds the face, the plate and the GPS position. If your phone backs up to a cloud service, a clear copy is probably there already. Gallery apps can keep the original too. Apple's Photos app on iPhone keeps it under every edit: open an edited photo, tap More and choose Revert to Original. With iCloud Photos those edits sync to every device on the account. A face pixelated inside the gallery app sits one tap away from coming back.

So share the file you download here, never the original. Don't send the clear version to a friend so they can hide the faces later. If the original should not exist at all, delete it yourself. Then check the cloud backup and any bin or recently deleted album your gallery keeps.


Turn a Photo Into Pixel Art

Pixel art is the other half of the job. The fun half. Choose Whole photo under What to Hide, keep Pixelate and the entire picture turns to blocks. Blocks across now counts across the photo's full width, from 4 to 512. The default 64 turns a 4,032 × 3,024 photo into 64 × 48 blocks and the readout says "64 blocks across · 63 px". An upright photo at the same setting gives 64 × 85.

How many blocks? That depends on what people should still see. Around 16 across, a photo becomes a patchwork of colour, more mood than picture. Around 32, faces turn into icons and a landscape into bands of sky, hills and ground. 64 suits most photos: clearly retro and clearly still your photo. From 128 upwards the blocks get fine enough to read as texture rather than style.

Rows follow by themselves. The tool picks however many rows keep each block as close to square as the photo's height allows, so you only ever set one number.

Some photos pixelate better than others. Strong, simple shapes survive. A red umbrella on a grey street, a single tree against the sky, a cat on a plain sofa. A busy crowd at 32 blocks across turns into noise. Size matters too. A subject that fills a third of the frame gets about 21 of the 64 blocks across. Crop in until it fills most of the frame and the same setting spends nearly every block on the subject. The Image Aspect Ratio Changer does that crop in a few seconds.

Retro Palettes

Each block normally keeps its own average colour. Pixel Art Colours swaps that for a fixed set, the way old hardware had to work. The choices are Original, PICO-8, Game Boy, CGA, Grayscale, 1-bit and From photo. The tool matches every block to the nearest colour in the set using OKLab, a colour space designed so that equal distances look about equally different to the eye. A plain distance in red, green and blue can pick a colour that is close in numbers and wrong to look at.

PaletteColoursValuesWhat it is
OriginalEvery block its ownNoneEach block keeps its own average colour.
PICO-816#000000 #1D2B53 #7E2553 #008751 #AB5236 #5F574F #C2C3C7 #FFF1E8 #FF004D #FFA300 #FFEC27 #00E436 #29ADFF #83769C #FF77A8 #FFCCAA16 colours of the PICO-8 fantasy console.
Game Boy4#0F380F #306230 #8BAC0F #9BBC0F4 greens, the usual stand-in for the original Game Boy screen.
CGA4#000000 #55FFFF #FF55FF #FFFFFFBlack, cyan, magenta and white, CGA's 4-colour mode.
Grayscale8#000000 #242424 #494949 #6D6D6D #929292 #B6B6B6 #DBDBDB #FFFFFF8 even steps from black to white.
1-bit2#000000 #FFFFFFBlack and white only.
From photo16Picked from each photoThe 16 colours the photo leans on most.

PICO-8 is a fantasy console from Lexaloffle Games, a made-up retro machine with a 128 × 128 pixel screen and 16 fixed colours. Its palette is bright and saturated. A peach, a pink and a brown sit among the blues and greens, so faces have somewhere to land.

The original Game Boy had no colour palette at all. Its screen showed four shades on a greenish display with no backlight. The 4 greens here are the approximation pixel artists settled on. There are no official RGB values to copy, because the hardware had none.

CGA, the Color Graphics Adapter, was IBM's first colour graphics card for the IBM PC. In palette 1 at high intensity it showed black, light cyan, light magenta and white. Four colours, none of them green or brown, so the matching gets bold. Run a beach photo through the tool's matcher and a clear sky snaps to cyan, pale sand to white and a sunburn to magenta. Green grass goes cyan as well.

Grayscale gives 8 even steps from black to white. 1-bit is black and white only, the harshest choice and often the most striking. From photo picks the 16 colours your photo leans on most. A beach stays sandy and a forest stays green while the blocks still snap to a tidy set.

Dither helps palettes with few colours. Instead of forcing every block to its nearest colour, it lets neighbouring blocks alternate between palette colours so the eye blends them into a shade the palette lacks. A sky in 1-bit without Dither is one flat slab. With Dither it fades from light to dark the way the real sky did. Dither works with Square and Circle blocks.

Sprites at One Pixel per Block

A pixelated photo at full size is still a big file: 4,032 × 3,024 pixels painted in 63 pixel squares. A game, a sticker set or a pattern wants the opposite. One pixel per block.

Saved Size offers exactly that when Whole photo, Square and Pixelate are all on. Full size W × H keeps every pixel. 1 px per block C × R saves a true sprite where each block becomes a single pixel. The default grid on a phone photo saves as a 64 × 48 image. Every pixel in it is one block's colour, with nothing resampled or guessed.

Retro targets fall out of the numbers. Slide to 128 blocks across and a landscape phone photo saves at 128 × 96, which fits inside the PICO-8's 128 × 128 screen. At 160 it saves at 160 × 120, the full width of the Game Boy's 160 × 144 screen. Add the matching palette and the photo looks native to the machine.

Small sprites look tiny on a modern screen. Scale them up in whole steps, 2×, 4× or 8×, with smoothing switched off in whatever program you use. Smoothing blurs the hard edges you just made. For a social post, save at Full size instead, so the blocks arrive already big and sharp.

Circle, Hexagon and Diamond Blocks

Square is the classic. The other three change the character of the result. Circle draws round dots, close to a finished bead piece or an LED sign. It also works with Dither. Hexagon builds a honeycomb that reads more like texture than a grid. Diamond stands every block on its point, like tiles laid on the diagonal.

The shapes work on marks too, so a censored face can wear honeycomb instead of squares. Pick on looks. The block count still decides how much stays hidden.

Two limits follow from geometry. Sprites need Square, because the pixels in a file are squares and a hexagon cannot be one pixel. Grid lines need Square too, because a chart needs straight rows to count along.

Grid Lines for Cross Stitch and Beads

Grid lines, with Whole photo and Square, draws a thin line between every block. The picture becomes a chart you can count: one block, one stitch or one bead.

Cross stitch first. 14-count Aida, the fabric most patterns are written for, carries 14 stitches to the inch. Divide the blocks across by 14 and you have the finished width in inches. Slide to 112 blocks across on a landscape phone photo and the chart comes out 112 × 84. That stitches to 8 × 6 inches. Pick a palette and each of its colours becomes one colour of floss. Game Boy needs 4 and PICO-8 needs 16.

Fuse beads next. A large square Perler pegboard holds 29 × 29 beads and four of them snap together into a 58 × 58 square. At 28 blocks across, a landscape photo charts at 28 × 21 and fits on one board. At 56 it charts at 56 × 42, fits on four boards and takes 2,352 beads. Standard beads measure 5 mm, so that piece comes out about 28 cm wide.

Save the chart as PNG and zoom in to count. Pattern software that reads one pixel per stitch wants the 1 px per block file instead.


Why the Colours Stay True

Averaging colours sounds trivial. Add up the pixels, divide, done. Do that on the numbers a file stores and the result comes out too dark.

The numbers in a JPG or a PNG are not amounts of light. They are gamma encoded, squeezed so that more of the scale goes to the dark tones where eyes notice small changes. Take a block that is half black and half white. Average the stored numbers and you get 128. But 128 gives off only 21.6% of white's light, well under the half you started with. The block turns murky.

This tool converts every pixel to linear light first, the way light really mixes. It averages there and converts back. The same half-and-half block comes out at 188, which carries 50% of white's light. That matches what you would see if you stepped back far enough for the two halves to blur together.

You notice it wherever a photo has fine contrast: striped shirts, lettering on a sign, bare branches against a bright sky, sunlight through a fence. Averaged the plain way, every one of those blocks darkens and the picture goes muddy where the detail was. Averaged in linear light, a pixelated sky stays as bright as the sky.

The grid is even too. With Blocks across, every block is within one pixel of every other, so no thin strip turns up at the right or bottom edge. At 128 blocks across a 4,032 pixel photo, blocks are 31 or 32 pixels wide. With Pixels, the tool centres the fixed blocks on the photo instead of starting at the top left. Choose 50 pixel blocks on that photo and you get 81 columns with a 41 pixel part block at each edge. A grid started at the left would leave a 32 pixel sliver down the right-hand side, the kind of seam an eye finds at once.

The download keeps the photo's own size. A 12.2 megapixel photo saves at 12.2 megapixels. The only limit is the device: 16.8 megapixels on iPhone and iPad before iOS 18, 33.6 from iOS 18 and 50 everywhere else. A bigger photo shrinks just enough to fit, with its shape unchanged.


Animated GIFs, Batches and File Formats

Animated GIFs keep every frame. Open one, mark the area and save with Save As set to Same as original. Every frame gets the same treatment. A screen recording with a password field in it comes out hidden from the first frame to the last. So does a reaction loop with a stranger walking through the back of it.

Marks stay put. They sit in one place on every frame and do not follow a face that moves, so cover the whole path. GIF frames hold 256 colours at most, which suits pixel art perfectly and photos less well. The tool also takes GIFs whose frames add up to 400,000,000 pixels in total. That is 3,086 frames of a 480 × 270 clip or 342 frames of a 1080 × 1080 one.

A batch takes up to 50 images. You share settings across the batch or set them per photo through the bar labelled Changes Apply To: All images, Only this image or Copy to all. Marks always belong to each photo. Auto find faces in all photos scans the whole batch in one go. Then step through and check each photo, the same as you would for one. Download N Images as ZIP saves the set as one archive. Every file keeps its own name with -pixelate-image-collagephotoapp.com added, so beach.jpg becomes beach-pixelate-image-collagephotoapp.com.jpg.

JPG, PNG, WebP, GIF, BMP and AVIF open in current browsers. HEIC opens wherever the browser can decode it, which the phone section covers. Save As offers Same as original, JPG, PNG and WebP. Same as original keeps your format with three exceptions: BMP and AVIF save as PNG. HEIC saves as JPG. JPG starts at quality 100, the top of the scale. WebP starts at 100, which Chrome writes as lossless WebP.

JobSave AsWhy
Pixel art, sprites and stitch or bead chartsPNGLossless, so every block stays one flat colour with a hard edge
A photo with hidden faces for social mediaSame as original or JPGSmaller files for photos. At quality 100 the soft edges JPG adds to blocks are faint
A screenshot with a hidden name or numberPNGScreen text and flat backgrounds stay crisp
Images for a websiteWebPAt quality 100, Chrome writes lossless WebP
An animated GIFSame as originalKeeps every frame. GIF holds 256 colours at most

Why not JPG for pixel art? JPG compression is lossy. It cuts the picture into 8 × 8 pixel blocks of its own. Its errors show most around sharp corners between contrasting colours. A pixelated image is nothing but sharp corners between contrasting colours.

The two grids rarely agree. The default grid on a phone photo has 63 column edges inside it. Only 7 of them land on a JPG block boundary, because 63 is not a multiple of 8. The rest fall inside JPG's blocks, where the compression smudges them a little. At quality 100 the smudge is faint. Save PNG and there is none.


Pixelating on a Phone

Every control works with a fingertip. A few habits make it easier.

Say you have a photo of your children on the swings, with three other people's children in the background. Open it and choose Marked areas. Tap Auto find faces. Delete the ovals over your own children with ×. Work across the far side of the playground for any face the finder missed, then draw an Oval over it. Press Original vs Pixelated, check the small Pixelated copy and press Download.

A thumb is clumsier than a mouse, so mark first and fix second. Draw the Oval roughly over the face, then tap it and drag its corners or edges until it covers the hair and the chin. Brush works well with a finger once Brush size is set. Moving the slider shows a circle of the new size in the middle of the photo. Original vs Pixelated lets you check the original underneath without taking anything off.

The squint test matters more on a phone, because a phone is where people will see your photo small. Check the thumbnail pair under Original vs Pixelated. Then hold the phone at arm's length and look at the preview itself as well. If the person comes back into focus at that distance, push the blocks bigger.

HEIC, the iPhone camera's format, opens in Safari. On an iPhone or iPad it also opens in Chrome, Edge and Firefox, because there they run on WebKit, the engine behind Safari. On Android and Windows those browsers can't decode HEIC, so save the photo as JPG first. On a Mac, open it in Safari.

Memory is the phone's other limit. iPhones and iPads before iOS 18 cap one picture at 16,777,216 pixels, about 16.8 megapixels. From iOS 18 the tool allows 33.6. A 12.2 megapixel photo fits under both with room to spare. A larger one saves scaled down just enough to fit, with its shape unchanged. Open the same file on a laptop and the full 50 megapixel allowance applies.


Privacy: Nothing Leaves Your Device

Everything is processed here itself, nothing is uploaded to our servers. The photo goes from your device into this page and back out to your downloads folder. No copy goes anywhere in between, so there is nothing for us to store, scan or lose.

That matters more for this tool than for most. People pixelate photos because they hold something private: a child's face, a home address, a bank statement. Sending that file to a stranger's server to hide it defeats the point. When we checked the pages ranking for this search in November 2026, several sent the photo to a server before a single block was drawn.

Auto find faces keeps the same promise. It runs on your device. The photo stays in the page and so do the boxes it finds.

Saved files carry no EXIF data and no GPS position, because each one is written fresh from pixels. There is no account to create and no watermark on the result. Close the tab and the photo is gone from the page.


From Pixelated Photo to Collage

A pixelated photo often ends up next to other photos: a school trip round-up, a party recap, a set of street shots. The Collage Maker lays several into one designed frame. Pixelate here first, then build the collage from the safe copies.

CollagePhotoApp, our free app for iPhone and Android, has two tools that suit this job. Its searchable emoji keyboard drops any emoji over a face, then lets you fade it, spin it or give it a shadow. Every photo in a collage also has its own blur among its editing tools. The store badges above and below this article link to both versions.

The other free tools pair well with this one. Add Text to Photo puts a caption on the finished image, such as "faces hidden for privacy". The Image Resizer shrinks a pixelated photo to a file size a form will accept. The Image Aspect Ratio Changer crops a crest or a street sign out of the frame before you pixelate what is left. Rotate Image stands a sideways phone photo upright. Merge Images joins photos in a row, a column or a grid. The Instagram Grid Maker cuts a pixel art piece into profile tiles that join up on your grid.

For collages made on a phone, read how to make a collage on your phone or the guide to making an Instagram Story collage. The free tools page lists every tool we make.


Frequently Asked Questions

How do I hide a password, card number or address in a screenshot?

Choose Marked areas, draw a Box over each private field and set the effect to Solid colour. Solid colour leaves no shapes behind. That matters for text, because a screenshot shows one known font on a flat background, the exact setup that pixelated-text attacks rely on. Cover the whole field, not just the middle digits. Then hunt for copies. The same card number can sit in an autofill dropdown, a notification banner or the title of a browser tab. An address often repeats in a map pin's label. One stray copy undoes the rest.

What should I cover before sharing a group chat screenshot?

Everyone who didn't agree to be in it. That means the names above each bubble, profile pictures, the phone numbers some apps show for people outside your contacts and every quoted reply, since a quote repeats an earlier message along with its author's name. Check the top bar too, where the group's own name and photo sit. A Box over each name does the job. Solid colour with Average fills each box with the average colour of what it covers, so the chat stays easy to read while the names disappear.

Is it safe to post a boarding pass once the barcode is pixelated?

Only if nothing readable survives. The barcode can hold more than the printed pass shows, including your name and the six-character booking reference. In 2020 a boarding pass photo that a former Australian prime minister posted on Instagram was enough to log in to his airline booking, which then exposed his passport number. So cover the whole code with Solid colour. A corner is not enough: a QR code can be built to scan with about 30% of it damaged. Fine blocks may leave it scannable too, since the code is made of squares already. Hide the printed booking reference as well.

What should I hide on an ID card or passport before sending a copy?

Only what the person asking doesn't need. Hide more and they may reject the copy. On a passport, don't miss the two lines of letters, numbers and < signs along the bottom of the photo page. That machine-readable zone encodes your name, document number, nationality, date of birth, sex and expiry date, so covering the printed number alone leaves it in plain view. Use Solid colour for numbers. Then use Add Text to Photo to write who the copy is for and the date across it. A copy stamped for one purpose is harder to reuse for another.

Can I pixelate a PDF?

Not directly. This page opens images: JPG, PNG, WebP, GIF, BMP and AVIF, plus HEIC where the browser can read it. For a PDF, screenshot the page or export it as an image, then mark what to hide. What you download is a flat picture with no text layer, so nothing under a box can be selected, searched or copied out. A rectangle drawn in a PDF editor can't promise that. In January 2019 reporters read the hidden parts of a court filing from Paul Manafort's defence team by copying the text under its black boxes and pasting it elsewhere.

Can a pixelated image still contain the original underneath?

Not one saved here. Every download is a fresh file built from the finished pixels, with no layers, no edit history and no leftover bytes from your original. Not every editor works that way. In 2023 researchers found that Markup, the screenshot editor on Google Pixel phones, saved a cropped image over the old file without trimming the file's end. The original's data stayed behind. The bug, nicknamed aCropalypse, let them rebuild the uncropped picture. Windows' Snipping Tool had a similar flaw.

How do I hide a licence plate in a car sale photo?

Choose Marked areas, drag a Box tight around the plate and pick Solid colour. With Average, the box takes on the plate's own blended colour, so it reads as a blank plate rather than a black bar. Pixelate works as well at the strong end of Block Size, 2 blocks across, where no character keeps its shape. Check every photo in the listing. A front shot and a rear shot both show a plate. Marks belong to each photo, so each needs its own. A parking permit on the windscreen can carry the plate number too.

How do I hide some children in a team photo but not others?

Choose Marked areas and tap Auto find faces, which puts a mark on every face it finds. Then remove the marks on the children whose families said yes: tap a mark and press × or Delete. Everyone still marked gets hidden. Check the back row closely. Our test photos included children half hidden behind adults and the finder boxed them all, yet a face turned away can still slip past. Name labels on bags and numbers on shirts deserve a look too, since a team list can match them to a child.

What should I hide in photos for a house sale or rental listing?

Anything that ties the home to the people living in it. The house number on the door or the letterbox. A car plate in the driveway. Family photos on the fridge, children's names on bedroom doors, post on the hall table and the calendar on the kitchen wall. Box handles flat things. Brush suits awkward shapes, such as a name stencilled along a wall. Up to 50 photos fit in one batch, each with its own marks. They download together as one ZIP. To show the rooms side by side, the Collage Maker fits a whole walkthrough on one page.

How do I hide strangers in the background of a holiday photo?

Choose Marked areas and tap Auto find faces. It marks everyone, you included, so tap the marks on yourself and your friends and press × to remove each one. The strangers stay covered and the scenery stays sharp. Small faces far back are where the finder struggles most, so look closely and draw an Oval over anyone it skipped. When the crowd fills the whole background, All except marked is quicker. Mark yourself and everything else gets hidden, scenery included.

Does pixelating a photo I already posted hide the person?

Not reliably. Once a picture has been online, anyone who saw it may have saved it, screenshotted it or shared it on. A pixelated repost can't reach those copies. Delete the original post first, then put up the pixelated version, which at least stops new viewers seeing the face. Next time, pixelate before the first post. That is the one moment you control every copy there is.

Why does All except marked hide my whole photo?

Because nothing is marked yet. All except marked hides everything outside the marks, so with no marks at all, the whole picture counts as outside. Draw a Box or an Oval over whoever should stay sharp. Auto find faces marks every face at once. The rest of the picture then pixelates around them. Marked areas behaves the other way round. With nothing marked it leaves the photo untouched, which is why a fresh batch in that mode looks unchanged until you start drawing.

Can pixelation be reversed?

Not back to the original pixels. Each block keeps one average colour and throws the rest away. What works is guessing and comparing. Depix, released in December 2020, reads pixelated text in screenshots by pixelating sample text in the same font and matching the blocks. Faces fall to matching too. In a 2016 study a neural network named the right person behind 16 × 16 pixel blocks 57% of the time among 530 people, 72% within its top five guesses. AI tools that claim to unpixelate a face invent the detail instead. In June 2020 one built on PULSE, from Duke University researchers, turned a pixelated Barack Obama into a white man.

Is a pixelated face anonymous enough for privacy law?

We can't tell you what your local law requires. Laws differ by country, so ask someone qualified where it matters. One benchmark helps, though. Europe's GDPR counts a person as identifiable if they can be identified "directly or indirectly", for example through location data or details of their physical or social identity. Research shows why indirect matters. People who know someone still recognise them through heavy pixelation, from hair, clothes and setting. A caption, a place tag or a uniform can name a person outright. Ask who could recognise them, not whether a stranger could.

Does Auto find faces recognise who is in the photo?

No. It detects faces, which is a different job from recognising them. The only question it answers is where a face is. For each one it returns a box and a score. There is no list of people behind it and no face prints. Recognition needs a second step that turns each face into numbers to compare against known people. This page has none.

Does Auto find faces upload my photo?

No. The scan runs on your device. Everything is processed here itself, nothing is uploaded to our servers. You can watch it happen. Open the network tab in your browser's developer tools, tap Auto find faces and you will see no photo leave.

Why did Auto find faces miss a face?

It misses more often on side profiles, faces turned away from the camera, faces in heavy shadow and very small faces at the back of a crowd. The finder already reads the whole photo plus overlapping tiles. When the faces are small, it reads smaller tiles again. Still, no face finder catches everything, so look before you save and draw an Oval over anyone it skipped. Hand-drawn marks and found ones behave the same: move, resize or delete either kind.

Why did Auto find faces put a mark where there is no face?

Now and then something face-like scores high enough to pass. In our test on six group photos holding 67 faces, the finder boxed all 67 and added 3 false boxes. An extra box costs little: a patch of wall or a sleeve gets pixelated for nothing. Tap it and press × or Delete. Of the two possible mistakes, that's the cheap one. A missed face is the one that matters, so check every result before you download.

Why is Auto find faces slow on some phones?

It runs on your device, so the device sets the pace. The finder uses the graphics chip where the browser allows it. Where it doesn't, the work falls back to the main processor, which is much slower. Big photos add work too. On a large picture the finder reads the whole frame plus overlapping tiles, then smaller tiles when the faces are small. Older phones feel that most. Closing other tabs frees memory. The scan still finishes, just more slowly.

How many faces can Auto find faces mark in one photo?

There is no set limit. Our busiest test photo held 21 faces and the finder boxed all 21. Each face gets an oval 40% wider and 65% taller than the detected face, so the hairline and ears go under the blocks too. In a tight group those ovals overlap. That does no harm, since the shared area stays hidden either way. For a whole set of club or class photos, Auto find faces in all photos runs the finder on every picture in the batch.

Does the Circle block shape let the photo show between the circles?

No. Every circle sits inside its own square block. The corners between circles take the same block colour in a darker shade, so not one original pixel shows through. That's what gives Circle its bead or LED sign look. Hexagon and Diamond leave no gaps to fill, because both shapes tile edge to edge. Pick the shape under Block shape. Dither works with Square and Circle only, since hexagons and diamonds don't line up in the rows and columns that dithering walks through.

What does Average do under Solid colour?

It fills each mark with that mark's own average colour, mixed the way light mixes. A box over a name on a white page turns pale grey. An oval over a face becomes a flat patch in that face's own tones. No shape survives, so it hides as completely as Black or White, yet it draws far less attention than a black bar. Use Average when you want the edit to go unnoticed. Use Black when you want readers to see that something was removed on purpose, as on a document.

Can I type my own emoji to cover a face?

Yes. Pick one of the 16 ready-made choices or type any emoji into Any emoji. Each mark gets the emoji drawn a little larger than the mark itself. Under it sits a coarse pixelate, because emoji have transparent corners that would otherwise show slivers of the face. Emoji works in Marked areas only. The design comes from your device's own emoji set, so an iPhone draws Apple's emoji and a Windows PC draws Microsoft's. Once saved, it's plain pixels. Everyone sees your version.

Can I use different effects on different faces in one photo?

Not in one pass. The effect covers every mark on a photo, so one face can't wear an emoji while another is pixelated. Two passes do it. Mark the first group, pick the first effect and save as PNG. Open that saved file, mark the rest and pick the second effect. PNG matters here, because it hands the second pass an exact copy of the first instead of a re-compressed one. In a batch, Only this image under Changes Apply To lets each photo keep its own effect.

What does Blur Strength measure?

The same scale as Block Size, on the same slider. Pick Blur and the slider takes the name Blur Strength, but right still means a stronger hide. Behind it, the blur radius grows with the block size that the same setting would give Pixelate. In Marked areas the blur also reads a little past each mark's edge, so the softened patch fades into its surroundings instead of stopping at a hard line. Pixelate's blocks stop exactly at the edge of the mark.

What is the difference between Blocks across and Pixels?

Blocks across sets how many blocks fill the width. Pixels sets how big each block is. Blocks across gives the same look on any size of photo and makes an even grid: on a 1,000 pixel wide image at 63 blocks, every block is 15 or 16 pixels wide. No thin strip is left at the edge. Pixels gives the same block size on every photo, which suits a batch of screenshots in mixed sizes. Its grid is centred, so the leftover splits between both edges. At 16 px on that same image, the first and last columns come out 12 px wide.

Should I crop a photo before or after pixelating it?

Before. In Whole photo the grid fits the width it is given, so a crop afterwards slices through blocks and leaves part blocks along the cut. Crop first in the Image Aspect Ratio Changer and the grid lands evenly on the final frame. Straighten first too. Level the horizon in Rotate Image before pixelating, since turning a pixelated picture tilts every block and softens its edges. For an avatar, a square crop first means a square sprite: a square photo at 32 blocks across saves a 32 × 32 image at 1 px per block.

Is it better to crop someone out than to pixelate them?

When they stand near the edge, yes. A crop removes a person completely, with no blocks, no outline and no clothing left to recognise. Pixelation always leaves a person-shaped patch where someone stood. So crop away whoever you can reach from the edges, then pixelate the people in the middle of the frame. A crop changes the frame's shape too. Pick a ratio that still fits where the photo is going, such as 4:5 for an Instagram feed post.

What is the difference between an 8-bit and a 16-bit pixel art look?

Mostly colour, not size. The 8-bit Nintendo Entertainment System drew 256 × 240 pixels and showed up to 25 colours at once out of 54. The 16-bit Super Nintendo drew 256 × 224 but showed up to 256 at once, picked from 32,768. Both screens were 256 pixels across, so 256 blocks across suits either look. For 8-bit, keep the colours few: PICO-8 with its 16 colours. Or a 4-colour set such as Game Boy or CGA. For 16-bit, leave Pixel Art Colours on Original or try From photo.

What does Dither do in pixel art mode?

Dither mixes palette colours across neighbouring blocks so the eye blends them into shades the palette lacks. Without it, every block snaps to its nearest colour and a smooth sky breaks into flat bands. With it, the tool carries each block's rounding error over to the blocks to its right and below, the Floyd-Steinberg method. The sky becomes a fine scatter of colours that reads as a gradient from a step back. It matters most with tiny palettes such as 1-bit, Game Boy and CGA. On Original there is no palette, so Dither has nothing to do.

How does From photo choose its colours?

It builds a palette out of the picture itself. The tool averages every block first. Then it sorts those block colours into 16 groups with a method called median cut, which keeps splitting the most varied group in half. Each group's average becomes one palette colour. Every block then takes the nearest of the 16, judged in OKLab, a colour space built to match how different two colours look to the eye. So the picture still looks like your photo, only with far fewer colours.

What size should pixel art be for a Discord emoji?

Discord suggests 128 × 128 pixels for a custom emoji and resizes other sizes itself. So choose the grid, not the pixels. At 128 pixels wide, 32 blocks across means 4 pixels per block. 16 blocks means 8, chunkier and easier to read at chat size. Make the photo square first with the Image Aspect Ratio Changer, so the blocks stay square when Discord crops it. Save as PNG, which keeps every block edge sharp and any transparent background clear.

Why can't I pick 1 px per block?

Because a sprite needs one grid of square blocks over the whole picture. Saved Size appears only with Whole photo, the Pixelate effect and Square blocks. Marked areas give every mark its own grid, so there is no single grid to shrink. A circle, a hexagon or a diamond can't shrink to one pixel and keep its shape. Animated GIFs always save at full size as well. Set those three options and the choice between Full size and 1 px per block appears. Grid lines follow a similar rule and need Whole photo with Square.

Does pixelating an image make the file smaller?

Pixelating the whole photo usually makes it much smaller. A file stores detail. Big flat blocks hold very little, so PNG in particular shrinks a lot. Marking a few faces changes far less. Most of the photo stays as it was and the tool saves JPG at quality 100, so the download can come out larger than the original if your camera saved it at a lower setting. For the smallest file, choose Saved Size 1 px per block: a 12.2 MP photo at 64 blocks across becomes a 64 × 48 image. For fewer pixels on an ordinary photo, use the Image Resizer.

What happens to transparent areas in a PNG?

They stay transparent if you save as PNG or WebP. Pixelate averages transparency along with colour, so a block that is half see-through comes out half see-through. A fully clear area stays clear. That keeps a logo or a sticker cut-out usable on any background. Solid colour is the exception: it fills a mark completely, so the mark itself turns opaque. JPG can't store transparency at all, so a JPG download lays everything on white. Same as original returns a PNG as a PNG.

Will the parts I didn't mark lose quality?

Barely. Only the save can touch them. The tool copies every unmarked pixel as it is. The one step after that is the encoder. JPG saves at quality 100 by default, the most detail a JPG can keep, though still not lossless. WebP goes out at 100, which Chrome writes losslessly. PNG never loses anything. Quality 100 is a default, not a lock. Lower it when a site caps upload sizes. Every step down saves bytes and softens fine detail a little more.

How are the pixelated files named?

The saved file keeps your original name and gains our suffix: holiday.jpg saves as holiday-pixelate-image-collagephotoapp.com.jpg. That is worth a second look on a privacy job. Cameras name files with numbers, but a file you renamed earlier, such as emma-school-play.jpg, carries the child's name into the download. Rename it before you share. The extension follows the format you save: BMP comes back as .png and HEIC as .jpg. Our name sits in the file name only, never in the picture.

Can I copy the pixelated image instead of downloading it?

Yes. Copy Image sends the finished picture to your clipboard as a PNG, so a chat, an email or a support ticket can take it straight from there. PNG means the copy loses nothing, see-through areas included. It's the quick route for a screenshot: cover the private details, then copy and paste without leaving a new file in your downloads. Browsers decide whether a page may write images to the clipboard, so the button works where yours allows it. Where it refuses, Download still works.

Can I pixelate an image from a web page without saving it first?

Yes, on a computer. Right-click the image, choose Copy image, then come back here and press Ctrl+V (Command+V on a Mac). The picture opens in the editor as if you had dropped in a file. Screenshots work the same way when your system copies them to the clipboard. Then mark what needs covering and use Copy Image to send the result on. Not a single file touches your downloads folder along the way.

How many images can I pixelate at once?

Up to 50 in one batch. Drop them all in and the download button counts them. With 12 photos it reads Download 12 Images as ZIP. One tap saves every picture in a single file. Settings can be shared or kept apart. With Changes Apply To on All images, one effect and block size cover the lot. Only this image changes the photo on screen alone. Copy to all hands its settings to every other photo. Each photo is opened, pixelated and closed in turn, so the batch never opens every photo at once.

Do marks drawn on one photo apply to the rest of a batch?

No. Marks belong to the photo you draw them on, because faces and plates sit in different places in every picture. Step through the batch and mark each one. Auto find faces in all photos does the first pass for you, then you check each photo in turn. Settings work differently. Effect, block size and palette follow the Changes Apply To bar, so the whole batch can share them or each photo can keep its own. A photo with no marks stays untouched in Marked areas.

Why won't my iPhone photo open on my Windows PC?

It's probably a HEIC file, the space-saving format iPhones use when Camera Formats is set to High Efficiency. HEIC opens here only in Safari and in the browsers on an iPhone or iPad, which all run on Safari's engine. Chrome, Edge and Firefox on Windows can't decode it. Three fixes. Open this page on the iPhone itself. Send the photo by AirDrop, Messages or email, which may convert it to JPG for a device that can't read HEIC. Or switch new photos to JPG under Settings > Camera > Formats > Most Compatible.

Does the pixelate tool work on Android phones?

Yes, in Chrome or Samsung Internet, with nothing to install. Tap to add photos from your gallery and draw marks with a fingertip. One catch on Samsung Galaxy phones: the Camera app's High efficiency pictures setting saves HEIF files. Samsung itself warns that some apps and websites don't support them. This page can't open them on Android either. Switch the setting off in the Camera app's settings and new photos save as JPG. Auto find faces uses the phone's graphics chip where the browser allows it.

Is there a limit to how large a photo I can pixelate?

Only what your device can safely hold. The download keeps the photo's own size up to 16.8 MP on an iPhone or iPad before iOS 18. iOS 18 and later allow 33.6 MP. Everywhere else the limit is 50 MP. A 48 MP iPhone photo measures 8064 × 6048, so it saves in full on a computer. Pixelated on the iPhone under iOS 18, it saves at 6688 × 5016, the same shape scaled down. Sprites skip the limit, since one pixel per block makes a tiny file. For every pixel of a huge photo, use a computer.

Can I pixelate a face in a video?

Not here. This page works on photos and animated GIFs and opens no video files, MP4 and MOV included. Hiding a face in a video takes a mark that follows it from frame to frame. Even in a GIF, the marks here stay in one place. Motion also makes faces easier to recognise. A study of famous faces in TV clips found moving faces were recognised better than still ones, even through heavy pixelation. For a short loop, turn the clip into a GIF first and draw a mark big enough for the face's whole path.

What happens when the face moves in an animated GIF?

The marks stay where you drew them on every frame. They don't follow the face. So watch the clip play first, note how far the face travels, then drag one Box over its whole range. When the subject crosses the frame, Whole photo is the safe choice. Leave Save As on Same as original to keep the animation. GIF frames hold 256 colours at most. Pixelated frames often fit inside that, so their colours survive exactly. One file can run to 2,500 frames at 400 × 400.

Can I pixelate images offline?

Not as an offline app. The page needs a connection to load. Some parts arrive only when first used. Once a part has loaded, the work happens on your device. Everything is processed here itself, nothing is uploaded to our servers. So a dropped connection never sends your photos anywhere. It can only stop a part that hasn't loaded yet from starting until you're back online.

Is the pixelate tool free with no watermark?

Yes. Every effect, shape, palette and batch is free, with no sign-up and no watermark. Nothing caps how many images you pixelate in a day. The practical limits are 50 images per batch and your device's memory. Our name appears in the file name only. A rename removes even that. Your own device does the work, so one more photo costs us nothing to serve.

Which keyboard shortcuts does the editor have?

A handful. Ctrl+V pastes an image in from anywhere on the page. The rest work once you click the preview. Ctrl+Z undoes a change to the marks. Ctrl+Y or Ctrl+Shift+Z brings it back. The arrow keys nudge the selected mark, Shift with an arrow moves it further and Delete removes it. On a Mac, Command takes Ctrl's place for undo, redo and paste.

What alt text should a pixelated photo get?

Describe the scene and say plainly that faces are hidden, without naming anyone you hid. "Under-10s team holding the cup, two faces pixelated" tells a screen reader user what sighted readers see. A name in the alt text, the caption or a tag undoes the pixelation for every reader, sighted or not. Skip "image of", since screen readers announce an image already. For a screenshot, give the gist and say what you covered: "Order confirmation, card number hidden". That text belongs to whoever posts the picture, in the platform's image description field.

Will the download match the preview?

Yes. The preview and the download go through the same drawing step with the same settings and the same marks. The preview draws a smaller copy sized for your screen, so it keeps up while you drag a mark around. The download goes back to your full photo and draws it again at its own size. Same blocks, same colours, more detail in the parts you left alone. Nothing saves until you press Download, so try every effect and palette first.

Make stunning posts from normal photos

Download CollagePhotoApp for free

GET IT ONGoogle PlayDownload on theApp Store