Reduce website image size without losing quality
A practical workflow for smaller website images without hurting visual trust.
Image Compressor
Compress images when possible — 100% free, 100% private
Drop images here or click to browse (up to 10)
JPEG, PNG, WebP, AVIF up to 50MB each
How It Works
Drop Your Images
Drag and drop up to 10 images at once, paste from clipboard, or click to browse. JPEG, PNG, WebP, AVIF up to 50MB each.
Compare & Adjust
Use the before/after slider to visually compare quality. Drag the quality slider to fine-tune.
Download
Download files individually, or grab all results as a single ZIP. File sizes are shown before you download.
Understanding Image Compression
How Lossy Compression Works
Lossy compression reduces file size by selectively discarding data that the human eye is least likely to notice. For JPEG, this involves the Discrete Cosine Transform (DCT), which converts spatial pixel data into frequency components. High-frequency details — subtle textures and minor color variations — are quantized more aggressively at lower quality settings. MozJPEG, the encoder Krunkit uses, applies optimized Huffman coding and trellis quantization, but the final output size always depends on the source image and settings.
Lossless Compression for PNG
PNG compression is always lossless — every pixel in the output is identical to the input. It works by applying prediction filters (Sub, Up, Average, Paeth) to each row of pixels, then compressing the filtered data with DEFLATE. OxiPNG, the Rust-based optimizer used by Krunkit, tests different filter and compression strategies to look for a smaller representation. Savings vary: screenshots, diagrams, and graphics with text often behave differently from already-optimized PNGs.
Choosing the Right Quality Setting
Quality settings don't map linearly to visual quality or output size. The "sweet spot" varies by image content: photos with smooth gradients may tolerate lower settings, while images with sharp text or fine lines often benefit from higher settings. For WebP and AVIF, browser support, encoder behavior, and the source image all affect the final balance between size and visible quality.
When to Compress vs. Convert
Compression keeps your existing format and reduces file size within it. Conversion changes the format entirely. If your audience uses modern browsers, converting JPEG to WebP or AVIF will yield larger savings than recompressing the JPEG. However, if compatibility matters — email attachments, print workflows, legacy systems — compressing within the original format is safer. For PNG images, lossless optimization is always worth doing before deciding whether to convert.
Frequently Asked Questions
How does image compression work?
Krunkit uses mozjpeg for JPEG, oxipng for PNG, and libwebp/libavif for WebP/AVIF. These are the same codecs used by Google, Facebook, and other major platforms. They remove redundant data and apply perceptually optimized compression to reduce file size while keeping the image looking great.
Will compression reduce my image quality?
At the default quality setting of 80, compression is virtually invisible to the human eye. You can drag the quality slider to compare — anything above 75 is typically indistinguishable from the original for photos.
What's the maximum file size I can compress?
Krunkit supports images up to 50MB. Since all processing happens in your browser, very large images may take a few seconds on slower devices.
Can I compress multiple images at once?
Yes! Drop up to 10 images at once, or paste from your clipboard. Each file is processed independently with its own status indicator. When done, download individually or as a single ZIP file.
Does compressing an image change its dimensions?
No. Compression only reduces file size by optimizing how the pixel data is encoded. Your image dimensions remain identical. Use the Resize tool if you need to change dimensions.
Related image optimization guides
WebP vs PNG vs JPG: which format should you use?
Choose the right output format before compressing or converting images.