Request | Modern Export Formats

Gigapixel AI Ideas
1

Hi - I’d like to see Gigapixel support AVIF and JXL. Reasoning shared below—same as in Photo AI!

2

Let’s take a close look at JXL.

Apple

  • Full support in iOS 17+, iPadOS 17+, macOS 14 Sonoma, and Safari 17+.
  • JPEG XL is also supported across Finder, Preview, Photos, Mail, Final Cut Pro, Pages, and other apps.

Samsung & JPEG XL?

  • Galaxy S24 Expert RAW now uses JPEG XL compression (wrapped in the DNG 1.7 container) instead of traditional RAW formats. This enables lossless or visually lossless compression for higher-quality captures with reduced file sizes.

Color Space & Fidelity Comparison

Format Bit Depth Support Color Space Support (RGB/CMYK/HDR) Wide Gamut (DCI-P3, Rec.2020) HDR (PQ/HLG) ICC Profiles / Metadata Notes for Archival Use
JPEG XL Up to 32-bit float Full RGB, CMYK, grayscale; HDR, XYB color model optimized for vision :check_mark: Full support (Rec.2020, P3, beyond) :check_mark: Native HDR10, PQ, HLG :check_mark: ICC profiles, Exif, XMP, IPTC, animation metadata Combines JPEG backward compatibility + modern wide-gamut HDR. Excellent future-proof archival format.
TIFF Up to 32-bit float RGB, CMYK, grayscale; widely flexible :check_mark: Supports wide gamut :check_mark: (via extensions, e.g. floating-point HDR TIFF) :check_mark: Robust ICC & metadata Industry archival standard; heavy files, no compression advantage vs. JXL.
PNG Up to 16-bit per channel Primarily RGB + alpha; grayscale Limited (not Rec.2020) ✘ No HDR :check_mark: ICC profiles Excellent lossless web graphics, not designed for HDR or wide gamut archival.
JPEG 2000 Up to 16-bit per channel RGB, CMYK, YCbCr, grayscale; HDR capable :check_mark: Supports Rec.2020 :check_mark: Supports HDR (PQ) :check_mark: ICC profiles, metadata Strong archival use in medical imaging, but less adopted outside niche areas.
HEIC/HEIF Up to 16-bit per channel RGB, YCbCr; grayscale; limited CMYK :check_mark: Supports wide gamut (Rec.2020) :check_mark: HDR10, PQ :check_mark: ICC & Exif metadata Efficient lossy/lossless container, but patent/licensing concerns.
WebP Up to 8-bit per channel RGB + alpha only ✘ Limited (sRGB focus) ✘ No HDR Limited ICC metadata Not suitable for archival; lacks wide gamut/HDR support.
AVIF Up to 12-bit per channel RGB, YCbCr; grayscale :check_mark: Supports Rec.2020, P3 :check_mark: HDR10, PQ, HLG :check_mark: ICC & Exif metadata Strong compression + HDR, but slower encode/decode vs. JXL.

Benefits of XYB used in the JXL standard

Benefit Why It Matters
Vision-optimized Errors are hidden where human vision is least sensitive (blue, high-frequency).
Better compression ratios Achieves JPEG-level sizes at higher fidelity, or smaller sizes at equal quality.
Smooth HDR & wide gamut Works naturally with Rec.2020 and HDR transfer functions.
No banding at low bitrates Perceptual quantization avoids the flat, posterized look of sRGB JPEG.
Efficient transforms Faster than more complex models like CIE Lab, suitable for real-time use.

Comparison to Other Models

  • sRGB: Simple device RGB, not perceptual; wastes bits in visually unimportant areas.

  • YCbCr (used in JPEG, HEVC, AVIF): Better than raw RGB, but not fully perceptual; tuned for broadcast video, not archival.

  • CIE Lab / ICtCp: Perceptually uniform, but mathematically heavier; slower for real-time image/video codecs.

  • XYB: Hits the sweet spot: close to perceptual uniformity + efficient math + compression-friendly.
    JPEG XL’s XYB color model, the broader concept it embodies is usually described as:

  • Perceptual Color Space → a color space tuned to human vision rather than device primaries.

  • Visual Psychophysics Optimization → mathematical modeling of how cones and opponent channels perceive differences.

  • Perceptual Quantization (PQ) → allocating bits more efficiently to luminance and chroma depending on visual sensitivity.

  • Opponent Color Coding → using opponent channels (like red–green, blue–yellow) to mirror how our eyes process color.

Why JPEG XL Has a Better Shot than the last challenger (JPEG2000)

JPEG XL was designed with lessons from JPEG 2000’s failure in mind:

  1. Royalty-free and open
  • Entire spec is free to implement, avoiding the patent mess.
  • Broad open-source libraries already exist (libjxl, ffmpeg support, etc.).
  1. Backward compatibility with JPEG
  • A JXL file can losslessly transcode existing JPEGs without recompression.
  • That means the entire JPEG corpus (trillions of images) could migrate with zero quality loss.
  • Huge advantage for adoption compared to JPEG 2000, which required re-encoding.
  1. Modern perceptual design (XYB color model)
  • Matches human vision better than YCbCr or wavelets.
  • Produces smaller files at higher quality than JPEG, WebP, and often AVIF.
  1. Wide feature set in one format
  • Lossless + lossy.
  • HDR + wide gamut.
  • Animation (like GIF/WebP).
  • Transparency (like PNG).
  • Progressive decoding + streaming.
  • Exif/ICC/XMP metadata support.
  • All in one spec — instead of juggling JPEG, PNG, GIF, WebP, AVIF.
  1. Performance optimized
  • Designed for modern CPUs/GPUs → faster to decode than JPEG 2000.
  • Encoding is heavier but improving rapidly with libjxl.
  1. Big-name adopters emerging
  • Apple (iOS 17, macOS 14, Safari).
  • Samsung Galaxy S24 Expert RAW (DNG 1.7 with JXL).
  • Adobe, Affinity, GIMP, Krita, etc.
  • Early signs of ecosystem traction, something JPEG 2000 lacked.

TIFF is 39 years old, which shows it has staying power. A standard for archivists.
JPEG is 33 years old. The most popular image standard in the world today.

If there is a transition to JXL, it will be selective.
Just like TV :television: didn’t replace the radio :radio:, and the radio didn’t replace the newspaper :rolled_up_newspaper:.
New media finds its own way to co-exist with it predecessors.

In addition to Apple’s adoption is Microsoft’s:
Microsoft

  • Support via the JPEG XL image extension on Windows 11 (24H2).
  • The Photos app gained native support in build 2025.11030.20006.0
3

Er, that’s a lot.

I do think there are interesting advantages with JXL as an archival format given its ability to do 32 bit lossless compression. I can’t find it, but I also read an interesting academic-ish article about using JXL as a format for some scientific work (astronomy).

As I outlined in my Photo AI ideas post, they’re both going to be important in the next gen image world.

If Gigapixel exports both PNG and JPG now, exporting as both AVIF and JXL (both of which have lossless options, notably), is a clear next step as we move forward with image formats.

4

I’m not going to help my argument about the urgency of need for AVIF/JXL export with this point here, but, it’s worth calling out:

I’m in the Apple ecosystem, and a fan. But statements about “full support” are… tenuous. Ever tried to turn a JXL 90 degrees in Preview? It wants you to turn it into a TIFF first. Photos can’t export JXL or AVIF. And so forth.

But. Formats move forward when people use them more. I would love to see Topaz helping to lead this, not waiting to see.

5

The ability to export in HEIC, not just JPG/PNG/TIFF, would be helpful for people relying on the native Apple Photos/iCloud ecosystem.

I’m an Apple ecosystem user. When I take an old image (usually a JPG or TIFF) and upscale/recover it, I’m either going to want to archive/print (TIFF) or put in a lossy format for family sharing.

While I continue to advocate for the ability to export JXL and AVIF, HEIC is going to be the most useful for any Apple ecosystem user right now. It’s far superior to JPG in terms of compression + image quality. Apple’s whole image ecosystem is built to happily handle HEIC.

Topaz Photo AI already can handle SDR HEIC exports—please add this to Gigapixel!

6

There is a format called Spectral JXL that some scientist are finding useful as a lossless format that allow them to use a DCT (Discrete Cosine Function) to make a more than visible spectrum into the new format.

Here in a more digestible format:

And this for the math buffs:
http://jcgt.org/published/0014/01/04/

7

I was disappointed to come away with the conclusion from Dakota’s many posts on where Topaz Studio is heading to learn that still the Apple will be the neglected step-child. I get it. If the gaming market was PC. Then the GPU market was PC. But the image processing market too?

I guess there are technical hurdles to work on the Silicon Apple hardware.

8

Ah, yes, this is the article I was thinking of. Thanks!

9

The information about Studio has come fairly fast to me as well. However, I’m unsure what you mean by “still the Apple will be the neglected step-child”… I haven’t seen anything regarding a downgrade of MacOS and Apple hardware support compared to other desktop support.

Still, I think that’s beyond the scope of this original thread about basic AVIF/JXL support.

You’ve got a complicated setup. It sounds like you’ve been really thoughtful about your direction with it. I think a new thread in the right forum (not “Ideas” and definitely not just AVIF/JXL support!) will probably get the right attention/help. Maybe Gigapixel "General’ to start and then someone at Topaz will hopefully either guide the thread to the right forum and/or be in direct touch.

10

I take back : " “still the Apple will be the neglected step-child”.

11

I mean, it’s just me and you here at the moment. :slight_smile: There is topic, yes, and there is also discussion. (Though your initial post about JXL was a… lot, I have to say.)

I hadn’t seen that post you linked to about Apple devices. That seems… problematic to me. I’m also responding there.

Again, losing the plot when it comes to AVIF/JXL support here…

12

The attraction of AVIF:

  • Superior Compression: AVIF compresses images much more efficiently than JPEG or PNG, resulting in drastically smaller files without compromise in visible quality.
  • High Color Depth and HDR: AVIF supports up to 12 bits per channel, enabling a wider dynamic range and richer color reproduction. It also supports HDR imaging and wide color gamuts for modern devices.
  • Transparency and Animation: Like PNG and GIF/WebP, AVIF supports transparency (alpha channel) and can store animated image sequences within a single file.
  • Versatile Metadata: AVIF can embed ICC color profiles, EXIF, XMP, geographic, and copyright metadata, making it suitable for professional, archival, and web use cases.
13

I don’t see JXL winning out there as a web standard.
Though, It may become potential archive medium.

14

As for AVIF, it seems there is a progression in time GIF–> WebP -->AVIF

GIF: June 1987 (GIF87a) (by CompuServe) - July 1989 (GIF89a)

WebP: Sept 2010 (by Google) - April 2018 (support library stable) - Nov 2024 RFC 9649

AVIF: Sept 2018 initial spec - Feb 2019 version 1.0.0-Jul 2020 ISO/IEC 23000-22

GIF took 2 years to stabilize
WebP took 14 years to become standards based
AVIF took 2 years to get to an RFC with the IETF.

AVIF, meanwhile has made solid progress to get an ISO/IEC standard.

1 Like
15

Here is an assessment of formats best for the web:

Format Strengths Weaknesses Suitability Browser Support
WebP Small size, transparency, animation, lossless & lossy No progressive loading Universal use All major browsers
AVIF Best compression, HDR, high bit-depth Slightly newer, less support on legacy devices Photos, premium sites Broad (2024+)
JPEG Legacy support, fast loading No transparency, lossy, larger files Photos All browsers
PNG Transparency, lossless Larger files Graphics All browsers

JPEG and PNG get full support. WebP is all the major browser. And AVIF is catching up.

AVIF has the better technology.

  • Better image quality.
  • Best compression.
  • Color space is excellent with high bit-depth.
2 Likes
16

I am just joining this topic. I would love Jpg-XL and AVIF support. But more urgently, I want Topaz photo to save files directly in an HEIC format that is compatible with Apple HEIC. Currently there are very few tools and browsers (including MS photo) that will open a Topaz-saved HEIC file. I realize this is because of the increased colour depth. However I use Topaz at the end of my pipeline and I want my production files in an HEIC format that is currently well supported. Right now I need to add an additional conversion at the end of my pipeline.

1 Like