Lossy or lossless: which compression should I use for web images?

Use lossy (JPG, WebP, AVIF) for photographs — the human eye barely notices the difference at quality 80–85, and file sizes are 5–20× smaller. Use lossless (PNG, WebP-lossless) for screenshots, UI mockups, logos and anything with sharp edges or text — lossy creates ugly artifacts around boundaries.

Does conversion preserve EXIF metadata (camera, GPS, date)?

It depends on the tool: ImageMagick and FFmpeg copy EXIF by default, while cwebp drops it unless you pass -metadata all . KaijuConverter strips metadata server-side after conversion to protect your privacy — if you need to keep camera or GPS data, use a desktop tool you control.

Lossy or lossless: which compression should I use for web images?

Use lossy (JPG, WebP, AVIF) for photographs — the human eye barely notices the difference at quality 80–85, and file sizes are 5–20× smaller. Use lossless (PNG, WebP-lossless) for screenshots, UI mockups, logos and anything with sharp edges or text — lossy creates ugly artifacts around boundaries.

Does conversion preserve EXIF metadata (camera, GPS, date)?

It depends on the tool: ImageMagick and FFmpeg copy EXIF by default, while cwebp drops it unless you pass -metadata all . KaijuConverter strips metadata server-side after conversion to protect your privacy — if you need to keep camera or GPS data, use a desktop tool you control.

Guide

JPEG XL (JXL): The Next-Generation Image Format

Q: What makes JPEG XL different from WebP and AVIF?

JPEG XL offers three unique advantages: (1) lossless JPEG transcoding — existing JPEGs can be stored as JXL with ~20% size reduction and then fully reconstructed to the original bit-for-bit JPEG; (2) progressive decoding — a low-quality preview is visible from the first bytes; (3) full HDR support with up to 32-bit float channels, wide-gamut color spaces, and ICC profiles. WebP and AVIF lack lossless JPEG transcoding and have limited HDR support.

Q: What makes JPEG XL diferente de WebP e AVIF?

JPEG XL oferece three unique advantages: (1) sem perdas JPEG transcoding — existing JPEGs can be stored as JXL com ~20% size reduction e then fully reconstructed para o original bit-for-bit JPEG; (2) progressive decoding — a baixa-quality preview is visible de the first bytes; (3) full HDR support com up to 32-bit float channels, wide-gamut color spaces, e ICC profiles. WebP e AVIF lack sem perdas JPEG transcoding e have limited HDR support.

Q: What makes JPEG XL anders als WebP und AVIF?

JPEG XL bietet three unique advantages: (1) verlustfrei JPEG transcoding — existing JPEGs can be stored as JXL mit ~20% size reduction und then fully reconstructed zum original bit-for-bit JPEG; (2) progressive decoding — a niedrig-quality preview is visible von the first bytes; (3) full HDR support mit up to 32-bit float channels, wide-gamut color spaces, und ICC profiles. WebP und AVIF lack verlustfrei JPEG transcoding und have limited HDR support.

PC By Pablo Cirre • Updated May 7, 2026

JPEG XL (JXL): The Next-Generation Image Format

What Is JPEG XL?

JPEG XL (extension .jxl) is a royalty-free, open image coding standard developed by the Joint Photographic Experts Group — the same body that created the original JPEG in 1992. Standardized as ISO/IEC 18181 between 2021 and 2022, JPEG XL was designed from the ground up to be a universal successor to both JPEG (for photographs) and PNG (for lossless and transparency), while also competing with modern formats like WebP, AVIF, and HEIC.

Its two defining properties:

Lossless JPEG recompression: An existing JPEG can be losslessly transcoded to JXL and reduced by roughly 20% in size, then fully reconstructed to the original bit-for-bit JPEG bytes. No other format offers this.
State-of-the-art compression: At equivalent quality to JPEG, JXL images are typically 30–60% smaller; at equivalent quality to PNG lossless, they are 15–35% smaller.

Technical Architecture

Modular Encoding

JXL uses a sophisticated modular prediction system (VarDCT mode for lossy, Modular mode for lossless) as its core coding engine.

VarDCT mode (lossy):

Adaptive block sizes from 8×8 to 256×256 (unlike JPEG's fixed 8×8)
Sophisticated edge-preserving quantization with adaptive Lagrange multipliers
XYB color space (perceptually uniform, derived from LMS cone responses) that allows more bits to go where the human eye is sensitive
Patches prediction, noise synthesis, and spline rendering for synthetic content
Progressive decoding: low-resolution preview from first bytes, refined as more data arrives

Modular mode (lossless and near-lossless):

MA (Meta-Adaptive) tree predictor using local image structure
Supports all bit depths: 1-bit line art through 32-bit HDR float
Delta palette for efficient encoding of graphics and UI screenshots

Bit Depth and Color

JXL natively handles:

1 to 32 bits per channel — binary, 8-bit, 10-bit, 12-bit, 16-bit, 32-bit float
Up to 4099 channels — wide-gamut, spectral imaging, multi-pass rendering layers
HDR and wide-gamut: full support for Rec. 2020, Display P3, Pro Photo RGB, and custom ICC profiles
Alpha transparency: full support, including premultiplied alpha and 16-bit alpha

Animation and Sequences

JXL supports multi-frame animation with:

Per-frame timing with millisecond precision
Blending modes (replace, alpha composite, multiply, etc.)
Frame reuse / delta frames for efficient animated content
Cropping regions per frame (store only changed pixels)

This positions JXL as a drop-in animated PNG (APNG) and GIF replacement with far superior compression.

Metadata

JXL embeds metadata using the same container as ISOBMFF:

Exif — camera metadata (shutter speed, GPS, orientation)
XMP — extensible metadata platform (copyright, keywords, color profiles)
ICC profiles — embedded color management (mandatory for HDR content)
JUMBF boxes — extensible JSON/binary metadata for C2PA provenance

JPEG Transcoding: The Killer Feature

No other modern image format can claim this: feed a .jpg file to the cjxl encoder in "lossless JPEG recompression" mode, and you get a .jxl file that:

Is ~20% smaller than the original JPEG
Can be decoded back to the exact original JPEG (same bytes, same EXIF, same chroma subsampling) using djxl --jpeg

This is possible because JXL stores the original DCT coefficients, quantization tables, and Huffman codes in a more efficient entropy-coded representation, without applying any new lossy transforms.

# Lossless JPEG → JXL (stores original DCT data)
cjxl input.jpg output.jxl --lossless_jpeg=1

# Restore the original JPEG bit-for-bit
djxl output.jxl restored.jpg --jpeg

# Lossy re-encode for maximum compression
cjxl input.jpg output.jxl -q 85   # quality 0–100 (like JPEG quality scale)

# Lossless PNG → JXL
cjxl input.png output.jxl -q 100

Compression Benchmarks

Typical file size at visually equivalent quality (vs. original JPEG at Q75):

Format	Size ratio	Notes
JPEG (original)	100%	Reference
WebP lossy	~70%	No progressive, limited bit depth
AVIF	~60–65%	Slow encode, no JPEG transcoding
JPEG XL	~50–60%	Progressive, JPEG transcoding, HDR
HEIC	~55–65%	Patent-encumbered, Apple-centric

For lossless (vs. PNG):

WebP lossless: ~70% of PNG
AVIF lossless: ~75–80% of PNG (variable)
JXL lossless: ~60–70% of PNG for photographs; ~80–90% for UI/screenshots (Modular mode's MA predictor isn't optimized for pixel-art palette graphics the way PNG is)

Browser Support Status

JXL has had a turbulent browser adoption path:

Chrome: Added behind a flag in 2021, then removed in Chrome 110 (2023) pending a formal spec finalization. As of 2025, the Chromium team has indicated plans to re-add support — active development is ongoing.
Firefox: Supported since Firefox 118 (2023) behind image.jxl.enabled flag; Nightly enables by default.
Safari: Supported in Safari 17.0+ (macOS Sonoma, iOS 17) — fully shipped. This makes JXL the default format on Apple devices.
Edge: No native support; follows Chromium.

The Safari shipping and Chrome's ongoing re-evaluation mean JXL is gaining real traction. For image delivery on the web today, serve JXL to Safari users and AVIF or WebP to Chrome/Firefox users.

Command-Line Tools

The reference implementation is libjxl (from the JPEG XL GitHub repository):

# Encode: JPEG → JXL (lossless JPEG recompression)
cjxl photo.jpg photo.jxl --lossless_jpeg=1

# Encode: PNG → JXL lossless
cjxl diagram.png diagram.jxl --lossless

# Encode: with effort level (1=fast, 9=slow+smaller)
cjxl photo.jpg photo.jxl -q 85 -e 7

# Decode to PNG
djxl photo.jxl photo.png

# Decode back to original JPEG
djxl photo.jxl photo.jpg --jpeg

# View metadata
jxlinfo photo.jxl

ImageMagick (7.1+), libvips, and FFmpeg (recent builds) also read/write JXL.

Use Cases Where JXL Excels

Photography archives: lossless JPEG transcoding — archive millions of existing JPEGs with 20% less storage, no quality loss, full JPEG recovery
HDR photography: 10/12/16-bit HDR images in Rec. 2020 or Display P3 with proper ICC profiles
Professional print: 32-bit float channels, wide-gamut, CMYK-compatible profiles
Animated content: better than APNG and WebP animated for high-quality animation
Progressive delivery: low-quality preview visible after ~1% of bytes; progressively refined to full quality
Scientific imaging: multi-channel spectral images, medical imaging, satellite imagery

Practical Tips

Use quality 85–90 for web photography (JXL quality is not identical to JPEG quality — calibrate visually)
Effort level 7 (-e 7) gives the best size/speed trade-off for production encoding; -e 9 is for archival
Always embed ICC profiles for HDR/wide-gamut content; browsers use them for correct color rendering
For web delivery today: detect Accept: image/jxl in request headers and serve JXL to Safari, AVIF/WebP to others
Progressive encoding is enabled by default for quality < 100 — no extra flag needed
JPEG transcoding is the immediate killer use case for large JPEG archives even before browser support is universal

JPEG XL's combination of lossless JPEG transcoding, best-in-class compression, true HDR support, and progressive delivery makes it the most technically ambitious image format ever standardized — and with Safari already shipping support, its real-world deployment is already underway.

Related conversions

Most teams that read this guide convert images in one of these directions:

Related conversions

Put what you just learned into practice — convert your files now in seconds, free and without registration.

Convert →

Convert →

Convert →

Convert →

Convert →

Convert →

Frequently Asked Questions

JPEG XL offers three unique advantages: (1) lossless JPEG transcoding — existing JPEGs can be stored as JXL with ~20% size reduction and then fully reconstructed to the original bit-for-bit JPEG; (2) progressive decoding — a low-quality preview is visible from the first bytes; (3) full HDR support with up to 32-bit float channels, wide-gamut color spaces, and ICC profiles. WebP and AVIF lack lossless JPEG transcoding and have limited HDR support.

JPEG XL oferece three unique advantages: (1) sem perdas JPEG transcoding — existing JPEGs can be stored as JXL com ~20% size reduction e then fully reconstructed para o original bit-for-bit JPEG; (2) progressive decoding — a baixa-quality preview is visible de the first bytes; (3) full HDR support com up to 32-bit float channels, wide-gamut color spaces, e ICC profiles. WebP e AVIF lack sem perdas JPEG transcoding e have limited HDR support.

JPEG XL bietet three unique advantages: (1) verlustfrei JPEG transcoding — existing JPEGs can be stored as JXL mit ~20% size reduction und then fully reconstructed zum original bit-for-bit JPEG; (2) progressive decoding — a niedrig-quality preview is visible von the first bytes; (3) full HDR support mit up to 32-bit float channels, wide-gamut color spaces, und ICC profiles. WebP und AVIF lack verlustfrei JPEG transcoding und have limited HDR support.

JPEG XL ofrece three unique advantages: (1) sin pérdidas JPEG transcoding — existing JPEGs can be stored as JXL con ~20% size reduction y then fully reconstructed al original bit-for-bit JPEG; (2) progressive decoding — a baja-quality preview is visible de the first bytes; (3) full HDR support con up to 32-bit float channels, wide-gamut color spaces, y ICC profiles. WebP y AVIF lack sin pérdidas JPEG transcoding y have limited HDR support.

Use <strong>lossy</strong> (JPG, WebP, AVIF) for photographs — the human eye barely notices the difference at quality 80–85, and file sizes are 5–20× smaller. Use <strong>lossless</strong> (PNG, WebP-lossless) for screenshots, UI mockups, logos and anything with sharp edges or text — lossy creates ugly artifacts around boundaries.

Safari 17+ (macOS Sonoma, iOS 17) fully supports JXL. Firefox supports it behind an experimental flag (image.jxl.enabled). Chrome removed its experimental JXL support in version 110 but the Chromium team has plans to re-add it. For web delivery today, detect the Accept: image/jxl request header and serve JXL to Safari users while sending AVIF or WebP to other browsers.

For new projects in 2026, WebP is supported by all modern browsers (95%+ of traffic) and saves 25–35% over JPG at the same visual quality. AVIF is even more efficient (40–50% smaller) but encoding is slower and Safari support is recent. Use WebP as the default and AVIF as the progressive enhancement via <code><picture></code> with JPG fallback.

When you run `cjxl input.jpg output.jxl --lossless_jpeg=1`, JXL stores the original JPEG DCT coefficients, quantization tables, and Huffman codes in a more efficient entropy-coded container — no new lossy transform is applied. The result is roughly 20% smaller. Running `djxl output.jxl restored.jpg --jpeg` reconstructs the exact original JPEG bytes, identical bit-for-bit.

When you run `cjxl input.jpg output.jxl --lossless_jpeg=1`, JXL stores the original JPEG DCT coefficients, quantization tables, e Huffman codes em um more efficient entropy-coded container — no new com perdas transform is applied. The result is roughly 20% smaller. Running `djxl output.jxl restored.jpg --jpeg` reconstructs the exact original JPEG bytes, identical bit-for-bit.

When you run `cjxl input.jpg output.jxl --lossless_jpeg=1`, JXL stores the original JPEG DCT coefficients, quantization tables, und Huffman codes in einem more efficient entropy-coded Container — no new verlustbehaftet transform is applied. The result is roughly 20% smaller. Running `djxl output.jxl restored.jpg --jpeg` reconstructs the exact original JPEG bytes, identical bit-for-bit.

When you run `cjxl input.jpg output.jxl --lossless_jpeg=1`, JXL stores the original JPEG DCT coefficients, quantization tables, y Huffman codes en un more efficient entropy-coded contenedor — no new con pérdidas transform is applied. The result is roughly 20% smaller. Running `djxl output.jxl restored.jpg --jpeg` reconstructs the exact original JPEG bytes, identical bit-for-bit.

It depends on the tool: ImageMagick and FFmpeg copy EXIF by default, while cwebp drops it unless you pass <code>-metadata all</code>. KaijuConverter strips metadata server-side after conversion to protect your privacy — if you need to keep camera or GPS data, use a desktop tool you control.

For web photography, quality 85–90 (out of 100) produces visually excellent results at roughly half the file size of JPEG Q75. Note that JXL quality values are not directly comparable to JPEG quality — calibrate visually for your content. Use effort level 7 (-e 7) for a good size/speed trade-off in production; effort 9 for archival maximum compression. For lossless, use -q 100 or --lossless.

For web photography, quality 85–90 (out of 100) produces visually excellent results at roughly half the tamanho do arquivo of JPEG Q75. Note that JXL quality values are not directly comparable to JPEG quality — calibrate visually para your content. usar effort level 7 (-e 7) para a bom size/speed trade-off in production; effort 9 para archival máximo compressão. para sem perdas, usar -q 100 ou --lossless.

For web photography, quality 85–90 (out von 100) produces visually excellent results at roughly half the Dateigröße von JPEG Q75. Note that JXL quality values are not directly comparable to JPEG quality — calibrate visually für your content. verwenden effort level 7 (-e 7) für a gut size/speed trade-off in production; effort 9 für archival maximal Komprimierung. für verlustfrei, verwenden -q 100 oder --lossless.

For web photography, quality 85–90 (out de 100) produces visually excellent results at roughly half the tamaño de archivo de JPEG Q75. Note that JXL quality values are not directly comparable to JPEG quality — calibrate visually para your content. usar effort level 7 (-e 7) para a bueno size/speed trade-off in production; effort 9 para archival máximo compresión. para sin pérdidas, usar -q 100 o --lossless.

Three common causes: (1) ICC color profiles dropped during conversion shift colors slightly; (2) chroma subsampling (4:2:0) reduces color accuracy in lossy JPG/WebP; (3) the new format may not support all features of the original (e.g. PNG → JPG drops transparency to white). For pixel-perfect results stick to lossless formats and preserve color profiles explicitly.