Image Compressor for JPG, PNG and WebP
Compress an image with adjustable output settings and dimensions. Compare the before-and-after file size, inspect the preview, and download the result.
Image Compressor workspace
Drop an image to compress
JPG, PNG, WebP, or GIF · up to 20 MB and 16,000,000 decoded pixels
Quality + Downscale
Fine-grained quality slider (1-100) plus optional max-width downscale. Combined they typically achieve 70-90% file-size reduction at imperceptible visual cost.
JPEG, WebP, PNG
Convert to JPEG (best for photos), WebP (~25-35% smaller than JPEG), or PNG (lossless — quality slider has no effect; use for graphics).
Debounced Before/After
See original vs compressed file size and percentage saved. Slider changes are debounced and only one encode runs at a time.
Bounded Client-Side
Compression uses Canvas or OffscreenCanvas in your browser, with file and decoded-pixel limits. No uploads, no signup.
Image Compressor: Shrink JPG, PNG & WebP in Your Browser
This client-side compressor re-encodes raster images through the browser Canvas API, letting you adjust quality sliders and maximum dimensions before downloading. Compression savings depend on the source image, output format, dimensions, and quality setting. A smaller result is not guaranteed, especially for an already optimized file. Compare the actual bytes and visible detail; PNG output does not use a lossy quality slider in the same way as JPEG or WebP.
For predictable browser memory use, files are limited to 20 MB, each side to 12,000 pixels, and decoded images to 16,000,000 pixels. Encoding is debounced, cancellable, and serialized to one active job.
How to use the image compressor
- Drag an image onto the drop zone, or click to pick a JPG, PNG, WebP, GIF, or AVIF file.
- Choose an output format: JPEG for photos, WebP for the smallest modern files, or PNG to stay lossless.
- Set quality with the slider (1–100) or a preset — High 92, Balanced 75, Compact 50, or Tiny 30. The slider has no effect on PNG. Changes are debounced before a new encode starts.
- Optionally drag max-width down to shrink the pixel dimensions; aspect ratio is preserved and the image is never upscaled.
- Read the before/after file size and percent saved, compare the side-by-side preview, then click Download to save the result.
How browser image compression actually works
Your browser decodes the source image, paints it onto a <canvas> or an OffscreenCanvas in a Worker when supported, and re-encodes the pixels with the native browser encoder. Browsers without the Worker path use the Canvas fallback. The quality argument is a number from 0 to 1 that only applies to lossy formats. As the MDN HTMLCanvasElement.toBlob() reference states, the quality value is used "if the requested type is image/jpeg or image/webp. If this argument is anything else, the default value for image quality is used." That is why this tool's slider does nothing for PNG — PNG is lossless, so the only lever left is dimension.
"A good Largest Contentful Paint score is 2.5 seconds or less" for at least 75% of page visits.— web.dev, Largest Contentful Paint (LCP). On most pages the LCP element is an image, so a heavy hero file is the fastest way to blow that budget.
Two lossy formats use distinct math. JPEG splits the image into 8×8 blocks and applies a discrete cosine transform (DCT), discarding high-frequency detail the eye barely notices. WebP uses VP8-based intra-frame prediction borrowed from video coding, which models smooth gradients and textures more efficiently — typically 25–35% smaller than JPEG at matching quality, with alpha transparency on top. PNG, by contrast, is always lossless: it stores every pixel exactly, so it is right for screenshots, logos, and text but oversized for photographs.
Worked examples: input → output
3.2 MB photo (4000px) · JPEG · quality 75 · max-width 1600
Downscaling to 1600px then re-encoding at 75% usually lands near 180–320 KB — roughly a 90% reduction with no artefacts visible at the display size.
Same photo · WebP · quality 75 · max-width 1600
WebP at the same settings typically comes out 25–35% smaller than the JPEG above — often around 120–220 KB — for visually identical output.
UI screenshot with text · PNG · max-width 1280
Keep PNG so text edges stay crisp. The quality slider is inert; shrinking max-width is the only way to cut bytes here. Converting to JPEG would blur the text.
Edge case · transparent PNG → JPEG
JPEG has no alpha channel, so this tool fills the canvas with the page background color before drawing (ctx.fillStyle = 'brand background color') — transparent areas become the page background color, not black, which is the usual failure of naive Canvas exports. If you need to keep transparency, output WebP or PNG instead.
Image Format Comparison (2026)
| Format | Best For | Transparency | Lossless | Typical Size |
|---|---|---|---|---|
| JPEG | Photographs, complex gradients, real-world imagery | No | No | Baseline (1.0×) |
| WebP (lossy) | Modern websites — replaces JPEG for photos and PNG for graphics | Yes (alpha) | Yes (separate mode) | ~0.70× of JPEG at same visual quality |
| AVIF | Next-gen photo compression; supported in modern browsers | Yes (alpha) | Yes | ~0.50× of JPEG; slower to encode |
| PNG | Screenshots, logos, line art, images needing transparency | Yes (alpha) | Always lossless | 5-10× of JPEG for photographs |
Default recommendation for new web content in 2026: WebP for photos and graphics, with PNG fallback only for very small icons where the format header would dominate.
How to Pick the Right Quality Level
| Quality Range | Visual Impact & Best Use |
|---|---|
| 95-100 | Nearly indistinguishable from source. Use when you cannot afford any visible artefacts. |
| 85-94 | Default for high-quality web photos. Artefacts only visible in side-by-side comparison. |
| 70-84 | Web standard for body images. Some artefacts visible in solid color areas; usually invisible at normal viewing. |
| 50-69 | Thumbnails, hero backgrounds, low-stakes images. Compression artefacts become noticeable. |
| 0-49 | Aggressive — use only for very small thumbnails or when bandwidth is critical. Artefacts are obvious. |
Always preview the compressed output at the target display size, not zoomed in — artefacts that are visible at 200% zoom may be invisible at the actual delivery resolution.
The white-fill trap most Canvas compressors miss
When you convert a transparent PNG to JPEG, JPEG cannot store an alpha channel, so the transparent pixels need a backing color. A naive Canvas export leaves them black. This tool paints the canvas with the brand page background color first (ctx.fillStyle = getBrandBackgroundColor(); ctx.fillRect(...)) before drawing the image, so a logo on a transparent background lands on the page background color rather than an ugly black box. If you need to keep transparency, choose WebP or PNG instead.
Two more behaviors worth knowing: the max-width slider only downscales — if you set a width larger than the source, the image stays at its native size (scale = maxWidth < naturalW ? ... : 1), so it never upscales and blurs. And every download is auto-named <original>-compressed.<ext>, so you never clobber your source file.
When and how to compress for web performance
Compress before you worry about quality: resize first. A 4000px photo shown at 800px is roughly 5× too large, so downscale to about 2× the maximum display width (for retina) using the max-width slider, then tune quality. For responsive layouts, export 1×, 2×, and 3× versions and let the browser pick with srcset rather than shipping one giant file to phones. Add loading="lazy" to every <img> below the fold so it does not block first paint, and let the heaviest above-the-fold image stay eager so it can win the LCP race.
Related Image & Web Tools
Multi-size PNG favicons from one image
Image to Base64Inline a compressed image in HTML/CSS
Base64 to ImageDecode a data URI back to a file
Aspect Ratio CalculatorPick a target width and height
CSS Filter GeneratorNon-destructive visual treatment
Open Graph GeneratorBuild social preview meta tags
Meta Tag PreviewerPreview your OG image in search and social
CSS MinifierShrink stylesheets for faster loads
JavaScript MinifierCut JS bytes to improve page speed
Guide: Open Graph Meta TagsSize and ship social preview images
AltCatch ExtensionVerify image alt attributes & HTTP status
All ToolsBrowse the full Toolk toolbox
Last updated: September 23, 2026 · Runs 100% in your browser — no uploads, tool input is not sent to Toolk.
Frequently asked questions
Does this image compressor require an account?
No. You can use the current tool without signup or an account. The visible file, pixel, and timing limits keep each job bounded in your browser.
Are my images uploaded to a server?
No. The entire pipeline — read, decode, re-encode, download — runs in your browser tab via Canvas or OffscreenCanvas and Blob URLs. Toolk’s page analytics do not receive your image bytes, no request carries file data, and nothing is stored. The image lives only in the current tab's memory and is cleared on refresh, so sensitive medical, personal, or internal images stay private.
How does the image compressor work?
Your browser decodes the image, draws it onto a Canvas or an OffscreenCanvas in a Worker when supported, then re-encodes it with the native browser image encoder. Browsers without that Worker path use the Canvas fallback. That Blob becomes the download. Everything is local: no server processing and no request carries the image bytes.
What size images are accepted?
Files are limited to 20 MB, each decoded side to 12,000 pixels, and the decoded image to 16,000,000 pixels. The limits are checked before compression so a pixel bomb cannot reach the encoder. Slider changes are debounced for 300 ms, and only one encode job is active at a time.
Why is the quality slider greyed out for PNG?
Because PNG is lossless, the quality argument to the browser encoder is ignored for image/png — only image/jpeg and image/webp honor it, per the WHATWG HTML encoding rules. For PNG, the only way to cut bytes is to shrink the pixel dimensions, so use the max-width downscale slider instead of quality. For photos, switch the output to JPEG or WebP to get real lossy savings.
Why does WebP compress better than JPEG?
WebP uses VP8-based intra-frame coding adapted from video compression, which handles smooth gradients and textures more efficiently than JPEG's DCT approach. At the same visual quality, WebP files are typically 25-35% smaller than JPEG and also support alpha transparency, so WebP can replace PNG for transparent graphics at a fraction of the size. Every major browser supports WebP natively in 2026.
Will my photo look worse at 75% quality?
Usually not. At 75% JPEG quality the difference from the source is typically imperceptible at normal viewing distance, even side by side. Artefacts appear at 50% and below, mostly in flat-color regions and around sharp edges. Use 75-85% for body images and 92% for hero images you expect people to inspect closely. The PNG slider does nothing because PNG is lossless.
Why is my compressed file LARGER than the original?
Re-encoding an already-compressed image can add bytes instead of removing them. Common causes: the source was already a low-quality JPEG you re-saved at higher quality; you converted a small PNG to JPEG above its effective quality; or the file was already tiny (under ~50 KB) where format overhead dominates. Start from the highest-quality original you have, or lower the quality slider until you see savings.
Does this strip EXIF metadata from my photos?
Yes. Browser Canvas and OffscreenCanvas re-encoding does not preserve EXIF, so the output keeps only pixel data — no camera make and model, GPS coordinates, timestamps, or custom tags. That is a privacy win when posting photos online. If you must keep EXIF (for asset-management systems that index by camera), use a tool with metadata pass-through such as mozjpeg --copy-all or ImageMagick without -strip.
Can I compress images in bulk?
Not here — this tool processes one image at a time to keep the byte comparison precise and the UI simple. For dozens or hundreds of files, use a CLI such as ImageMagick, mozjpeg, or sharp via a Node.js script. The same Canvas-to-Blob pipeline can be looped programmatically if you want to replicate this exact algorithm.
Why is there no AVIF output option?
Because the browser encoder does not emit AVIF in most browsers as of 2026, this tool outputs JPEG, WebP, and PNG only. AVIF is typically 30-50% smaller than WebP at the same quality with full alpha and HDR support, and browser decoding is now universal. To encode AVIF, use an external tool like squoosh.app, avif.io, or the avifenc CLI.