Unpixelate Image: Smooth Out Pixelated Photos
Enlarge a small or blocky picture and smooth away the jagged squares, with a before/after slider to judge the result. Nothing is uploaded.
Runs in your browser · nothing is uploaded
No image yet — choose, drop or paste one above to see the preview.
Original: — → Result: —
Your image is processed on your device and never uploaded.
What unpixelating can and cannot do
A pixelated image is one where individual pixels are large enough to see — usually because a small picture has been enlarged, or because heavy compression has turned smooth areas into blocks. This tool depixelates in three passes: optional edge-preserving smoothing on the original pixels to calm compression noise, a high-quality bicubic upscale by 2×, 3× or 4× done in gentle steps, and finally a soft blur followed by an unsharp mask, so edges look smooth but still defined.
To be clear about expectations: no tool can recover detail that was never captured. This is classic image processing, not AI. It will not rebuild a face from a dozen pixels or make a blurred number plate readable, and any service that claims to reverse a deliberate mosaic is inventing the result. What it does well is make small images look clean and natural at a larger size instead of blocky.
Like every tool on this site, it runs entirely in your browser. Your image is not uploaded, there is no account or watermark, and you can use it as often as you like. Results can be up to 8,000 × 8,000 pixels.
How to unpixelate an image
- Choose, drop or paste the pixelated image.
- Pick an upscale factor: 2× for mild pixelation, 3× or 4× for tiny images such as avatars and icons.
- Drag the before/after slider across the preview. Raise Smoothing until the squares disappear, then add Sharpen to bring edges back.
- If the original has JPEG blocks or colour noise, raise Edge-preserving smoothing: it flattens noise inside areas while leaving outlines alone.
- Choose PNG for the cleanest result, or JPEG/WebP for a smaller file, and click Download image.
Good jobs for this tool
- Enlarging a tiny avatar or profile picture. A 96 × 96 or 128 × 128 image saved from an old forum, chat app or game becomes a usable 384 × 384 or 512 × 512 picture without hard squares.
- Old low-resolution photos. Pictures from early digital cameras and phones (640 × 480, or about 1 megapixel) look much better at 2× with smoothing and light sharpening.
- Screenshots and thumbnails. A screenshot taken in a small window, or a thumbnail that is the only copy you have, can be enlarged for a slide, document or blog post.
- Small logos and icons for a quick mock-up. When you only have a small raster logo, 4× with moderate smoothing gives a clean placeholder. For anything final, ask for the vector (SVG or PDF) original.
- Over-compressed JPEGs. Edge-preserving smoothing reduces the 8 × 8 blocking and colour speckles that appear when an image has been saved or shared too many times.
How the processing works
Enlarging a small image in a photo viewer often uses nearest-neighbour scaling, which copies each pixel into a bigger square — that is what produces the staircase look. Bicubic resampling instead estimates every new pixel from a 4 × 4 neighbourhood with a smooth curve. This tool uses the Mitchell–Netravali filter, a standard choice that balances sharpness against smoothness, and enlarges in steps of at most 2× so the curve stays gentle.
After upscaling, a small Gaussian blur, scaled to the enlargement factor, removes the remaining grid pattern. The unsharp mask then adds contrast only along edges, so outlines look crisp without bringing the blocks back. The edge-preserving option is a bilateral filter: it averages neighbouring pixels only when their colours are similar, so it smooths flat areas but stops at edges.
Some pictures have already been enlarged into visible squares — a screenshot of a zoomed-in image, for example. In that case the tool detects the regular block grid, averages each block back to a single pixel and interpolates between the block centres with a Catmull-Rom spline (a bicubic curve that passes exactly through each centre), which removes the squares far more cleanly than blurring them would.
| Method | What it does | Typical look |
|---|---|---|
| Nearest neighbour | Copies each pixel into a larger square | Hard, visible blocks |
| Bilinear | Blends the four nearest pixels | Soft, with jagged diagonals |
| Bicubic + smoothing + sharpening (this tool) | Smooth curve over 16 pixels, then blur and unsharp mask | Clean, smooth edges without blocks |
| AI super-resolution | A trained model guesses plausible detail | Sharper, but may invent textures or features that were not there |
Tips for the best result
- Start with the largest version of the image you can find. Unpixelating a 400-pixel copy gives far better results than a 100-pixel copy of the same picture.
- Use the lowest factor that gives you the size you need. You can stretch or resize the result to an exact pixel size afterwards.
- Balance smoothing and sharpening. Too much smoothing looks waxy; too much sharpening creates bright halos along edges. Park the slider over a face or a line of text to judge it.
- For pixel art, where crisp squares are the whole point, do not use this tool — enlarge with nearest-neighbour scaling instead.
- If the image is also dark, clean it up here first and then lift it with the image brightener. Brightening amplifies noise, so it helps to smooth first.
- Save as PNG if you plan to edit further. JPEG adds its own compression blocks, which is exactly what you are trying to remove.
Limits and honest expectations
Unpixelating improves how an image looks; it does not add information. Text that is unreadable in the original will be smoother but still unreadable, and a face made of a few pixels will stay vague. Pixelation applied on purpose to hide faces or numbers cannot be reversed by this tool or by any honest software.
The result is limited to 8,000 × 8,000 pixels, so a factor that would exceed this is blocked. Processing a large result takes a few seconds and a lot of memory, and phones may struggle with outputs above about 20 megapixels. When the result is large, the preview is computed at a reduced size, so the downloaded file can look slightly crisper than the preview.
Most browsers cannot open HEIC photos, so convert them to JPEG first. Transparent PNGs keep their transparency when you export as PNG or WebP.
Frequently asked questions
Can you really unpixelate an image?
Partly. You can remove the blocky look by upscaling with smooth interpolation and then blurring and sharpening, which is what this tool does. You cannot restore detail that was never captured, and no honest tool can reverse a deliberate mosaic over a face or text.
How do I fix a pixelated picture for free?
Load the image on this page, choose 2×, 3× or 4×, adjust smoothing and sharpening while comparing with the slider, and download. There is no sign-up, watermark or limit, and the image never leaves your device.
Is this the same as AI upscaling?
No. AI upscalers use trained neural networks that guess new detail, which can look impressive but may invent features. This tool uses predictable image processing — bicubic resampling, Gaussian blur, an unsharp mask and a bilateral filter — that only works with the pixels you have.
Which upscale factor should I choose?
Choose the smallest factor that reaches the size you need: 2× for mildly pixelated photos, 3× or 4× for very small images such as 64–128 pixel avatars. Larger factors give smoother but softer results.
Is depixelating the same as unpixelating?
Yes. Depixelate, unpixelate and “fix a pixelated image” all describe the same task: making visible pixel blocks disappear so the picture looks smooth.
Are my images uploaded?
No. All processing happens in your browser with the Canvas API. Nothing is sent to a server, so private photos stay on your device.