Image Pixelator — Brush-Paint Pixelation, Censor Faces & Make Pixel Art

What an image pixelator does — and why the brush is the star

An image pixelator replaces groups of pixels with a single average colour, creating a blocky mosaic. People use it for two different jobs: to hide something (a face, licence plate, address, or text) or to make something look retro (pixel art, 8-bit avatars, album covers).

Most pixelators force you to choose between pixelating everything or drawing rectangular boxes. This one adds a third option that is often faster and more natural: a paintbrush. Drag over a face, a logo, a name badge, a password field, a licence plate — only that area gets pixelated. You control the shape with your hand, not with a box. Brush mosaic is the default tool for that reason: for most people it is the quickest route from “I need to hide this” to “done”.

Everything runs locally. The file you drop is decoded by your browser with the FileReader API, drawn to a canvas, painted on, and handed back as a new file. No server receives the bytes, and the tool keeps working offline once the page has loaded.

How to paint pixelation with the brush

1. Load the image. Drag a file onto the upload zone, click it to browse, or paste from the clipboard with Ctrl+V.
2. Brush mosaic is already selected. The brush toolbar appears above the canvas with a size slider and a Paint / Eraser toggle.
3. Adjust the brush size with the slider or the [ / ] keyboard shortcuts. A circle preview follows your pointer so you can see exactly where the pixelation will land.
4. Drag over the image to paint pixelation. Every stroke is recorded — you can undo individual strokes with Ctrl+Z or remove them all at once with Clear strokes.
5. Switch to Eraser with the Eraser button or the E key to remove pixelation you painted by mistake.
6. Fine-tune block size and palette from the controls on the right. The brush strokes stay in place; only the appearance of the pixelation changes.

💡 Tip: for a face, use a brush slightly larger than the face and paint over the whole head. It is better to over-cover than to leave an identifying ear, jawline or hair pattern visible.

The four tools compared

How pixelation works under the hood

Pixelation is a three-step operation, and understanding it explains why some block sizes work and others don’t:

1. Downsample. The image is divided into a grid of blockSize × blockSize squares. Each square becomes one pixel whose colour is the average of every pixel inside it.
2. Quantise. If a palette is active, each averaged colour is snapped to the nearest colour in that palette. Dithering diffuses the error into neighbouring blocks.
3. Upscale. The tiny image is scaled back up with nearest-neighbour interpolation — every pixel becomes a solid square of blockSize × blockSize.

The brush adds a fourth step: a mask. Each brush stroke paints into a mask canvas. During the final composite, the pixelated version is only drawn where the mask is white; everywhere else the (filtered) original image shows through. That is why brush strokes feel instant — the expensive pixelation work happens once, and the mask only controls which parts of it are visible.

Pixelation is not redaction — the Depix problem

Pixelation of text can be reversed. The open-source Depix project and follow-up tools like Unredacter demonstrated that pixelated passwords and card numbers can be reconstructed by rendering candidate characters through the same pixelation filter and matching the blocks. Averages leak: if the hidden content comes from a small alphabet (digits, letters), an attacker can brute-force the possibilities.

Check the rest of the image too: a reflection in a window, the URL bar in a screenshot, the filename in a document title — all can give away what you tried to hide. This tool exports a fresh canvas, so EXIF metadata (GPS, camera model) is stripped, but other clues in the image itself remain your responsibility.

Choosing the right block size

A useful sanity check: after pixelating, zoom the browser to 50% and look away for two seconds, then look back. If you can still tell what the hidden thing was, the block size (or brush coverage) is too small.

The retro palettes explained

• Game Boy (4 colours) — #0f380f, #306230, #8bac0f, #9bbc0f. Almost any photo becomes instantly nostalgic.
• PICO-8 (16 colours) — including #1d2b53, #ff004d, #fff1e8. Famously hard to make an ugly image with.
• Commodore 64 (16 colours) — the C64's muddy browns and blues, including #68372b and #6c5eb5.
• NES (16-colour subset) — a balanced slice covering black, white, greys, blues, reds, greens and yellows.
• Grayscale — luminance only, for newspaper-style pixel art or checking tonal structure.
• Sepia — a warm monochrome tint. Pairs well with 8 px blocks for vintage photos.
• High contrast — posterises to a handful of tones. Good for screen-print-style graphics.

Dithering matters: without it, a 4-colour Game Boy palette produces flat bands that look crude on a photograph. With Floyd–Steinberg error diffusion the eye blends the four greens into what looks like continuous shading — the same trick original Game Boy artists used to make 4 colours feel like 16.

How to pixelate a face in a photo

1. Load the photo and pick the Hide a Face preset, or just stay in brush mode with 20 px blocks.
2. Adjust the brush so it is slightly larger than the face. A single brush stroke is easier than multiple small ones.
3. Paint over the whole head — hair, ears, jawline. A face that is 90% covered is still recognisable from the 10% that is left.
4. Increase the block size to 24 or 32 if anything at all is still readable.
5. For anything more sensitive than a casual group photo, use Solid fill instead. A black rectangle is definitively unreadable; a pixelated face is not.

How to make pixel art from a photo

1. Load a photo with clear shapes — portraits, landscapes and objects work better than busy crowds.
2. Click the Pixel Art preset (8–12 px blocks, PICO-8 palette, Floyd–Steinberg dithering).
3. Adjust brightness and contrast until shadow and highlight separate cleanly.
4. Set Size to Pixel art in the footer — this exports one pixel per block.
5. Download the PNG and open it in Aseprite, Piskel or LibreSprite. Hand-tune: clean up stray pixels, sharpen outlines, add highlights. Automated pixelation gives you a base; the polish is your work.

What people use this for

• Privacy before publishing. Redacting faces in event photos, number plates in car listings, names on badges, account numbers in screenshots.
• Bug reports and tutorials. Hiding customer data, internal URLs, API keys and Slack messages before posting a screenshot publicly.
• Pixel art and game assets. Generating a base image to trace, or producing placeholder art in a consistent retro style.
• Social media and branding. Avatars, banners and album covers with a deliberate low-fi aesthetic.
• Meme edits. The “pixelated for comedic effect” look — a quick brush stroke over a surprised face is instantly recognisable.
• Print and textile design. Heavy pixelation (48–96 px blocks) creates large colour blocks that work as cross-stitch patterns or printed mosaic tiles.

Keyboard shortcuts

Limitations

• GIF and animated formats. Only the first frame of an animated GIF or animated WebP is processed.
• Very large images. Files above roughly 6000 px on the longest side take a second or two on older phones. The spinner shows while work is happening and renders are debounced.
• EXIF orientation. Modern browsers apply the image's EXIF orientation when drawing to canvas, so portrait phone photos appear upright. Very old browsers may ignore it.
• No secure redaction guarantee. Pixelation is not cryptography. Do not treat a pixelated image as a guarantee that the original content is unrecoverable — use solid fill for anything that matters.
• Colour profiles. The canvas works in sRGB. Images with an embedded non-sRGB profile (Adobe RGB, Display P3) are converted to sRGB, which can shift saturated colours slightly.
• No batch mode. One image at a time. Download your result before loading another.

Common mistakes to avoid

• Using pixelation for passwords. The single most dangerous mistake. Use solid fill.
• Painting too small an area. A brush stroke over only the eyes leaves the jawline, hair and ear visible — often enough to identify a person. Cover the whole head.
• Choosing a block size that is too small. A 4 px block over a face leaves it perfectly recognisable. Err large.
• Pixelating one copy and leaking another. If you pixelate a screenshot to post publicly but the original is still in your drafts folder, the edit accomplished nothing.
• Forgetting the surrounding context. The URL bar in a browser screenshot, the file path in a code editor, the timestamp in a photo corner — all leak information pixelation did not cover.
• Assuming the exported file is smaller. A pixelated PNG at original resolution is usually larger than the source JPG because PNG is lossless. If file size matters, export JPG or use the pixel-art output size.

Frequently asked questions

Related image tools on AIToolsPros