SDR: white is the ceiling
A sheet of paper, a cloud and the sun all end up at the same value, because there is nothing above white to put them at. Whatever the difference was in the original scene is gone by the time it reaches you.
Find out what your device can do, and see why HDR is not simply making everything brighter. Your own screen does the demonstrating.
Start the HDR testTwo short tests. In each one you will see a list of bars and click the bar where something changes — you do not need to be exact, and if two bars next to each other look the same, either is a fine answer. The first test asks whether your screen can show anything brighter than an ordinary white background. The second asks how much further above that it can still show a real difference in brightness. Neither test depends on telling colours apart. Neither gives you a percentage, either: you get a plain number in stops above reference white (and what that is in nits), not a claim about your screen's spec-sheet peak brightness.
Checking what this browser can send…
Checking what this browser can send…
A workshop with a window, rendered rather than photographed, so every value in it is a stated quantity of light rather than whatever a camera happened to record.
Rendering the scene…
Only one version is shown here
This browser has no way to put HDR on the page, so there is no HDR half to compare against. What is above is the tone-mapped SDR version on its own. Showing it twice and labelling one of them HDR is exactly the misleading comparison this page is about, so it is not done.
The highlight map still works: it is an ordinary image, and it shows which parts of the scene an HDR display would have put above white.
Both versions are the same scene, rendered from the same light. The SDR version is not darkened, desaturated or flattened: everything below 183 cd/m² is untouched, and only the highlights above it are compressed to fit. Colour is preserved exactly — the curve is applied to all three channels together.
| Part of the scene | Luminance | Relative to white |
|---|---|---|
| Shadow under the bench | 0.2 cd/m² | 1/1015 |
| Workshop wall | 6 cd/m² | 1/34 |
| Sheet of paper | 203 cd/m² | 1.0x |
| Sunlit bench | 320 cd/m² | 1.6x |
| Sky through the window | 1600 cd/m² | 7.9x |
| Chrome highlight | 2500 cd/m² | 12x |
| Edge of the sun | 7000 cd/m² | 34x |
A rendered scene can state its own light exactly, which is why the comparison above uses one. This is the trade for a genuine photograph instead: real, extreme contrast, put through the same tone-map and encoder, but honest that its brightness is recovered rather than stated.
Only one version is shown here
This browser has no way to put HDR on the page, so there is no HDR half to compare against. What is above is the tone-mapped SDR version on its own.
Both halves are the same photograph, put through this page's own tone-map — the same code the rendered scene above uses, not a separate demonstration.
Mount Tamalpais, from the OpenEXR project's sample images (ScanLines/MtTamWest.exr, © Industrial Light & Magic, BSD-3-Clause licence). Roughly 26.5% of the frame sits above reference white, peaking around 2.1 stops above it — an exposure choice made when preparing this page, not a value the photograph states on its own. See scripts/generate-hdr-photo.mts for exactly how.
HDR gives a picture more room between its darkest part and its brightest part. It does not make everything brighter.
A sheet of paper, a cloud and the sun all end up at the same value, because there is nothing above white to put them at. Whatever the difference was in the original scene is gone by the time it reaches you.
Ordinary white stays at a normal, comfortable brightness. The room above it is for things that are actually very bright, like the sun. The paper stays where it was. The sun does not.
That is why turning the brightness up is not the same thing. Brightness moves the whole picture, including the shadows. Range moves only the top.
Luminance is measured in cd/m², candelas per square metre, usually called nits. It is an amount of light coming off a surface, so it is a physical quantity rather than a setting.
A typical laptop in a room might put white at 100 to 300 cd/m². HDR content is normally mastered with reference white at 203 and highlights reaching 1,000 or more.
The ratio between the darkest thing a display can show and the brightest. Both ends count: a display that reaches 1,500 cd/m² but cannot render a dark grey has a worse range than the number suggests, which is most of why OLED and Mini-LED sets look different from each other on the same content.
The pale segment is the black level — how far off true black the darkest pixel actually sits. The second set has a lower peak and a longer usable range above its own floor, which is why the bigger number does not automatically win.
One stop is a doubling of light. It is the unit photographers already think in, and it makes headroom legible: 100, 200, 400, 800, 1,600 is four stops.
Each bar is the last one doubled — equal steps in stops, not equal steps in cd/m².
Reference white to a 1,600 cd/m² highlight is about three stops. That sounds small and is not: it is eight times the light.
Fitting content into a display with less range than the content has. Every HDR display does it, because content is routinely mastered brighter than anything can show, and how well it is done is much of what separates two sets showing the same film.
The SDR half of the comparison above is tone-mapped, and how is stated there rather than left to be trusted.
Gamut is which colours a display can show; dynamic range is how much light. They travel together in practice — HDR formats specify wide-gamut primaries — but they are independent, and a wide-gamut display is not necessarily an HDR one. The capability table above reports them separately for that reason.
You cannot create HDR on an SDR display, and this does not try to. It is a drawing of the idea in ordinary greys: the slider changes how much room the simulated display has above white, and everything below white deliberately does not move.
With no headroom, every highlight lands on exactly the same white as the paper. That is SDR: white is the ceiling, and a bulb and a sheet of paper are the same value.
The scale is logarithmic: each label is roughly ten times the light of the one before it. The marked band is what the simulated display can reach.
A screenshot of an HDR image is not a reliable way to show someone HDR. It can work, and increasingly does — several platforms now capture, store and display HDR end to end. What it cannot do is work dependably, because it only survives if every stage of the chain preserves it.
Any one of those can convert to SDR, and most do it silently. The person sharing the screenshot sees it correctly on the device that took it, which is exactly why the practice persists.
Try it on this page
Take a screenshot of the comparison above and open it beside the original. If your whole chain preserves HDR you will see no difference. If it does not, you have just demonstrated the point, and neither result tells you anything about the display — only about the pipeline between it and the file.
A vivid HDR image beside a grey, washed-out one labelled SDR.
SDR does not look like that. A properly tone-mapped SDR version of the same picture keeps its contrast, its colour and its shadows, and gives up only the brightness of the highlights. The comparison on this page is generated that way, and says what it did.
Two versions of an image, one uniformly brighter than the other.
That is a demonstration of brightness, which is a setting, not a demonstration of dynamic range. HDR lets one part of a picture be far brighter while the rest stays where it was. If everything moved together, nothing has been shown.
A screenshot of an HDR image, posted as proof of what HDR looks like.
The picture has to survive capture, storage, upload, whatever the platform does to it, the viewing app, the operating system and finally the display. Any one of those stages can convert it to SDR, and most of them are invisible to the person posting it.
| Format | Metadata | Typical use | Strength | Weakness |
|---|---|---|---|---|
| Absolute | HEIC/HEIF photos, some AVIF, and this page’s own PNGs | A code value means a specific quantity of light, the same on every display. | Defined far past what any screen can show, so almost everything still needs mapping down. | |
| Relative | HEIC photos from some cameras, alongside PQ | The bottom of its curve matches ordinary SDR, so a viewer that ignores it still shows something sensible. | Its brightness scales to the display, so the same file is not guaranteed to look the same on two HDR screens. | |
| A second image, not a curve | Camera photos from Android and iPhone, in JPEG or HEIC | The file a plain viewer opens is an ordinary, correctly-exposed SDR photo — nothing breaks for anyone. | The extra brightness is relative to that SDR rendering and the viewer’s own headroom, not a stated amount of light. |
Select a format for more on how it works.
Deliberately generic. Naming models and their capabilities means maintaining that list continuously, and a device table that is quietly two years out of date is worse than none — it is wrong with an air of authority.
Drop in a file of your own and this reads its metadata for the formats above: whether it declares PQ, HLG or a gain map, its bit depth and dimensions, and anything else its container states. Then it puts the file's own bytes next to a version forced through a plain SDR canvas, so you can see what its declared HDR signal actually adds on this browser and this screen.
Read in this tab only. Nothing here is uploaded anywhere, and it is gone the moment you remove it, choose another file, or leave this page.