How much smaller can a photo get without visible quality loss?
Phone photos and camera JPEGs commonly shrink by 60% to 90% at the default Balanced (0.75) preset, and even further when you cap Largest side to 1920 px or 1280 px to reduce image file size. As noted in the MDN HTMLCanvasElement.toBlob() Reference, toBlob() accepts a 0.0 to 1.0 quality argument for lossy image/jpeg and image/webp encoding when you compress photos online.
If you need exact custom pixel dimensions rather than a longest-side cap, you can scale images to exact pixel widths and heights before or after compressing.
Choosing Smaller file, Balanced, or Best quality presets
The Quality dropdown offers three tuned presets: Smaller file (0.6), Balanced (0.75), and Best quality (0.88). Balanced removes imperceptible high-frequency color noise while keeping textures, skin tones, and text edges clean on retina screens.
For JPG files, lossy encoders apply ITU-T T.81 DCT quantization matrices across 8×8 pixel blocks to discard high-frequency noise that human eyes barely notice.
Selecting Smaller file (0.6) pushes quantization slightly further for strict email or portal limits, while Best quality (0.88) preserves subtle gradients in product photography and portraits. Most users leave Balanced (0.75) active for everyday web uploads.
How the automatic 0.62 retry and no-bloat guard work
If a lossy export at 0.75 or 0.88 comes out larger than the input file (common with already-optimized web JPEGs), the compressor automatically retries once at 0.62. If the output at Keep size and Same as the original is still larger or equal, the no-bloat guard returns your original file bytes untouched.
Comparing WebP, JPG, and lossless PNG compression
Choosing the right output Format matters as much as the Quality slider in any image compression format comparison. Under IETF RFC 9649 WebP Image Format Specification, WebP combines VP8 lossy compression with optional alpha transparency, making it ideal for both photos and transparent UI screenshots.
| Format setting | Compression type | Alpha transparency | Quality slider active | Best use case |
|---|---|---|---|---|
| Same as the original (JPG/WebP) | Lossy | WebP only | Yes (0.6, 0.75, 0.88) | Quick shrinking without changing extensions |
| Same as the original (PNG) | Lossless Deflate | Preserved (32-bit RGBA) | Disabled (lossless) | Pixel-exact icons and lossless archives |
| WebP (smallest for websites) | Lossy VP8 + alpha | Preserved | Yes (0.6, 0.75, 0.88) | Fast web pages and transparent screenshots |
| JPG (works everywhere) | Lossy DCT | Filled with #ffffff |
Yes (0.6, 0.75, 0.88) | Email attachments and strict upload forms |
When converting transparent PNGs to JPG, see-through areas are filled with solid white (#ffffff) because JPEG has no alpha channel. If you ever need to reverse a WebP back into an uncompressed graphic for desktop editing, use a lossless WebP to PNG converter.
Choosing WebP (smallest for websites) keeps alpha transparency intact while applying lossy VP8 compression to RGB channels, making it ideal for UI screenshots with drop shadows or product cutouts that would otherwise balloon in lossless PNG.
Halving downscale steps with the Largest side selector
Resizing oversized camera originals is the fastest way to slash megabytes without pushing lossy compression into blocky artifacts. The Largest side dropdown lets you choose Keep size, 3840 px (4K), 2560 px, 1920 px (Full HD), or 1280 px (web, email).
Picking 3840 px, 2560 px, 1920 px, or 1280 px caps
For email attachments, CMS blog posts, and portal uploads, 1280 px or 1920 px provides sharp full-screen clarity at a fraction of a 4000 px raw photo’s byte size. Keep 2560 px or 3840 px for high-resolution hero banners or digital displays.
Images whose longest edge is already smaller than your chosen Largest side cap are never upscaled (Math.min(1, ...)), so mixed batches of large camera photos and small thumbnails can share one setting safely.
Why progressive halving keeps resized edges crisp
Instead of shrinking a 4000 px photo down to 1280 px in one coarse jump, drawResized() steps down in halves (cw / 2 >= targetW) with imageSmoothingQuality = "high" before the final draw. That progressive halving prevents jagged moire patterns on fine fabrics, brickwork, and UI text.
According to the W3C Portable Network Graphics (PNG) Specification Third Edition, PNG always uses lossless Deflate compression, so lowering Largest side is the only way to shrink a PNG without switching to WebP or JPG. If you are starting from modern .avif files for an older portal, you can turn modern AVIF shots into universal JPGs directly.
Local browser encoding and EXIF metadata privacy
Every image is decoded, resized, and encoded inside your browser tab using local memory, never uploaded to an external compression server. Images larger than 4096 px on a side or 16,777,216 total pixels are clamped proportionally to stay within browser 2D canvas limits.
Orientation correction without remote uploads
As specified in the MDN createImageBitmap() Web API Reference, decoding with imageOrientation: "from-image" rotates portrait phone photos upright before resizing or encoding.
Tip: If you drop a folder of PNG screenshots and see the Quality dropdown greyed out, switch Format to WebP (smallest for websites) to shrink them by 70% or more while keeping sharp text.
When re-encoding strips EXIF versus keeping original bytes
Re-encoding or resizing via canvas automatically strips camera EXIF tags and GPS coordinates from the -compressed output. When the no-bloat guard hands back your untouched original file, you can strip EXIF and GPS tags without re-encoding separately.