Skip to main content

Pixel Crunch 2.0

Your images. Your device. Your tools.

Pixel Crunch brings together focused image tools that work without sending your files to a processing API.

Code and resources are downloaded from Pixel Crunch; your images remain in the browser while they are processed.

Choose a tool

01

Compress

Reduce JPG, PNG, WebP, GIF, and SVG file size with quality control and batch downloads.

Compress images
02

Convert

Convert browser-compatible images to JPG, PNG, WebP, or AVIF.

Convert images
03

Remove background

Remove backgrounds locally with a model loaded only after you start processing.

Remove background

Practical privacy

No remote processing API receives your images.

One URL per task

Share or bookmark the exact tool you need.

Focused loading

Each page hydrates only its active tool.

How it works

From file to result, without intermediaries

Each tool follows a short workflow and keeps the work inside your browser session.

  1. 01

    Choose a task

    Open the tool you need and select one or more compatible images.

  2. 02

    Process on your device

    The browser reads and transforms the files locally. Pixel Crunch does not send them to a processing API.

  3. 03

    Review and download

    Inspect the result, adjust available options, and save individual files or a ZIP when supported.

Technology and data flow

Cloudflare delivers the application. Your browser processes the image.

Pixel Crunch is a static site. There is no Pixel Crunch API that receives the image to compress, convert, or analyze it.

01 · Delivery

Cloudflare Pages

Serves Pixel Crunch HTML, styles, JavaScript, and other public resources.

02 · Processing

Your browser

Reads the image from your device and performs the transformation in memory with web APIs.

03 · Result

Your download

Creates a local Blob and saves it to your device as an individual file or ZIP.

What crosses the network

The application and its public resources are downloaded. The image you select is not attached to a processing request. When removing a background, the model is downloaded to the browser before local inference runs.

File API

Reads selected files without sending them through a remote form.

Web Workers

Move compatible tasks off the main thread to keep the interface available.

Canvas + codecs

The browser decodes and re-encodes the formats it actually supports.

Blob + ZIP

Builds downloads in memory and groups batch results when supported.

What each tool does

Use the right tool for each image

Compression, conversion, and background removal solve different problems while keeping processing in your browser.

01

Compress images

Reduce file size while aiming to preserve useful visual quality. Suitable for websites, email, and storage.

Open compressor
  • JPG, PNG, WebP, GIF, and SVG
  • Quality controls and comparison
  • Batch processing and ZIP download

02

Convert images

Change format for compatibility, transparency, or size. Availability depends on your browser codecs.

Open converter
  • JPG, PNG, WebP, GIF, and AVIF input
  • JPG, PNG, WebP, or AVIF output
  • Animated GIF: exports the first frame

03

Remove background

A model running on your device separates the subject and creates transparent output without a remote inference API.

Remove background
  • JPG, PNG, or WebP input
  • Model loads only after you start
  • Transparent PNG or WebP output

Verifiable privacy

Your images are not part of a processing request

Cloudflare Pages delivers HTML, CSS, JavaScript, and public resources. Transformation runs in the browser, and Pixel Crunch has no endpoint that receives your images.

Review the source code (GitHub)

No account or history

You do not need to register, and Pixel Crunch does not maintain a library of your files.

Ephemeral files

Images stay in memory during the session and can be cleared from the tool.

Open source

The project and its AGPL-3.0-only license are publicly reviewable.

Clear network boundaries

Your browser downloads the application and the public model when removing a background. That is different from uploading an image.

Quick format guide

The right format depends on the content

Changing format does not improve an image by itself. Choose based on quality, transparency, animation, and compatibility.

JPG

Photography and broad compatibility.

Uses lossy compression and does not support transparency.

PNG

Logos, screenshots, and graphics with transparency.

Can be heavy for photographs.

WebP

Web images with a useful quality-to-size balance.

Output depends on browser support.

AVIF

Reducing compatible images in modern browsers.

Encoding and support are not uniform across browsers.

GIF

Simple animation and compatibility.

The converter exports only the first frame.

SVG

Logos and vector illustrations.

Optimized by the compressor; not accepted by the raster converter.

Frequently asked questions

What to know before you start

Does Pixel Crunch upload my images?

Not to a processing server. The application transforms files in the browser. Your browser still requests the site, scripts, and other public resources required to run it.

What is the difference between compression and conversion?

Compression aims to reduce file size while keeping the format or purpose. Conversion creates a different format to change compatibility, transparency, or efficiency.

Can image quality change?

Yes. Lossy formats and color reduction can alter detail or color. The compressor lets you adjust and compare before saving.

Is there a file size limit?

The current interface accepts files up to 10 MB. The practical number and size of images also depend on available device memory.

Does it work offline?

First-visit and complete offline use are not guaranteed. You must download the application first; background removal also needs to download its model before local inference runs.

How does background removal work?

The browser downloads the model, analyzes the image on your device, and creates transparent output. The image is not sent to a remote inference API; speed and available memory depend on the device.