How do glowing HDR images work?

A gain-map file contains two images: the untouched sRGB image, plus a hidden grayscale glow map marking which pixels may exceed the screen's normal white and by how much. HDR displays apply the map; everything else ignores it and shows a normal image. It is the ISO 21496-1 gain map format, the same mechanism iPhone photos have used since iOS 17.

Which displays show the glow?

Screens with HDR brightness headroom: MacBook Pro 2021 and later glows strongly, other Mac laptops from around 2016 glow gently when dimmed, iPhone 12 and later, most OLED phones, and true HDR monitors and TVs. Standard-range screens show correct colors with no glow. For the strongest glow, set screen brightness to about 50%.

Which apps and browsers show the glow?

Chrome and Edge on HDR displays, Android, and iOS 17+ Photos render gain-map glow. macOS Preview shows correct colors without the boost. Anywhere unsupported, the image simply looks normal.

Do glowing images change the original colors?

No. The base image keeps the original sRGB pixels, byte-for-byte for JPEG input. The glow travels as separate metadata, so viewers that do not support it show the image exactly as authored.

Why do glowing LinkedIn company logos work?

LinkedIn preserves PNG color profiles on company-page logos, so a PNG converted to Rec.2020 PQ with a BT.2100 PQ profile keeps its glow there. Pipelines that re-encode images, like Slack, strip gain maps (leaving a normal image) and can render PQ PNGs washed out, which is why the gain-map JPEG is the safer default.

image-luminescence

Your images, glowing.
Colors untouched.

Whites lift beyond your screen’s white on HDR displays. Everything else stays exactly as you made it.

Drop an image or click to choose

png or jpeg · jpeg bases kept byte-for-byte

Set your screen brightness to about 50% — in Chrome on an HDR display the two converted versions below keep their whites lit while everything else dims.

Watercolor disco ball, original — original

Luminesced — glows on HDR displays — gain-map jpeg · the default

PQ version — glows where color profiles survive — pq png · for pipelines like linkedin

One file, two pictures.

SDR screens see the first picture. HDR screens see the file become the second. Nobody sees both.

Everyone sees

drop or click

HDR reveals

drop or click

mode

The third image below is one file built from the other two — what it shows depends on the screen reading it. Thumbnails always show the first. Reveals need the gain map to survive, so re-encoding pipelines (LinkedIn, Slack) show only the first picture — share the file directly or host it yourself.

Reveal ingredient A: the disco ball — everyone sees

One file that shows A on SDR displays and B on HDR displays — the file · same bytes

One file, two images.

Your image, untouched. The file’s base is your exact sRGB pixels — byte-for-byte when you feed it a jpeg. Every display on earth shows your true colors.

A glow map rides along. A second, hidden grayscale image — this is the disco ball’s, for real. White pixels mean “boost the matching pixel,” black means “leave it alone.” Nothing in your image is blurred or changed.

HDR displays apply it. Capable screens boost the marked pixels into brightness headroom. Everything else ignores the stencil and sees a perfectly normal jpeg.

The curve behind the sliders: nothing is boosted until a pixel nears white, then the glow ramps to full strength. The format is the ISO 21496-1 gain map — the same mechanism every iPhone photo has used since iOS 17.

Take it home.

command line

# it's a homebrew package:
brew tap kwicz/tap
brew install image-luminescence

luminesce logo.png
# -> logo-luminescence.jpg
luminesce-reveal a.png b.png
luminesce-pq logo.png  # pq png, for linkedin

--boost 49.26  glow ceiling, default max
--knee  0.85   how near white before glow

Prefer source? git clone the repo, brew install imagemagick libultrahdr, and run the python scripts directly — same tools, no install.

which tool

Two converters, two strategies. luminescence.py makes gain-map jpegs: glow as removable metadata, safe everywhere — pipelines that re-encode (Slack) just show a normal image. Use it by default. luminescence_pq.py makes PQ pngs: glow baked into the pixels, survives pipelines that keep PNG profiles (LinkedIn — how those glowing company icons work), but viewers that ignore the profile show a washed-out ghost (Slack). Only for destinations you’ve verified. image-reveal (luminescence_reveal.py) packs two different pictures into one file: SDR viewers see the first, HDR viewers see the file become the second. One artifact of the format nobody else is using.

this page

The converter above runs the same math compiled to the browser — canvas in, gain-map jpeg out. Your image never leaves your machine.

roadmap

Also: brew tap kwicz/tap && brew install image-luminescence, pip install from the clone, a Figma plugin, and a Finder quick action. Details in the repo.

displays

The glow needs a screen with brightness headroom. Rough vintages: MacBook Pro 2021+ glows hard, other Macs from ~2016 glow gently when dimmed, iPhone 12+ and OLED phones glow, and HDR monitors/TVs (roughly 2016+, real HDR — not “HDR400”) glow. Standard-range screens like Apple’s Thunderbolt Display show correct colors, no glow. Headroom shrinks as brightness rises — set your screen to about 50% brightness for maximum glow. This page checks your display live, in the corner below.

format

ISO 21496-1 / Ultra HDR gain-map jpeg · glows in Chrome and Edge on HDR displays, Android, and iOS 17+ Photos. macOS Preview shows correct colors but may not boost. Anywhere unsupported it’s simply a normal image — correct colors, no glow. The gain curve — how hard each pixel is boosted — ramps from your threshold to your ceiling and is stored as metadata, which is why macOS describes these files as “Adaptive Gain Curve.”