What real HDR logos actually encode
Anyone can claim a file is HDR. We decoded a set of HDR logo files and measured them — colour signalling, peak luminance, and how much of the frame is driven above reference white.
These are measured examples of real HDR glow files, with the numbers behind each one. Every image here was produced by this free online tool and then read back: encoded peak luminance in nits, the share of the frame above the 203-nit diffuse white reference, and the signalling each file carries — cICP 9 16 0 1 for PNG, or a Rec.2100 PQ ICC profile for JPEG. Nothing is a mock-up or an illustration of the effect. On a display with HDR headroom these files genuinely render brighter than the page around them.
The headline finding
Every file in this set is signalled the same way — BT.2020 primaries with the PQ transfer function, a cICP tuple of 9 16 0 1. That part is textbook.
What they do with it is not. 5 of 6 drive their white to PQ code 255 — the top of the encoding, nominally 10,000 nits. That is the naive approach: attach a PQ profile and let white burn as hard as the format allows, without deciding what the highlight should actually be.
5 of 6
encode white at the PQ maximum
10,000 nits requested
18.5%
median share above reference white
the typical bright area
91.6%
worst case in this set
nearly the whole frame driven above white
Why driving everything to maximum is a mistake
Displays reach their highest peak brightness on small areas. Push most of the frame above reference white and the panel’s power limiting tone-maps the whole image back down — so an image asking for 10,000 nits across 91.6% of its area will typically look duller than one asking for 1,600 across 5%.
There is also no headroom left for intent. If white is already at the ceiling, a second element cannot be brighter than it, and nothing can be emphasised relative to anything else.
The measurements
Peak is the highest absolute luminance any pixel encodes. Above-white is the share of the frame sitting beyond the 203-nit reference level.
| File | Size | Signalling | Peak | Above white |
|---|---|---|---|---|
| real-hdr-portrait.jpg | 800×800 | BT.2020 · PQ9 16 0 1 | 10,000 nits | 91.6% |
| taboola_logo.jpeg | 400×400 | BT.2020 · PQ9 16 0 1 | 10,000 nits | 36.5% |
| cloudfive_net_logo.jpeg | 400×400 | BT.2020 · PQ9 16 0 1 | 10,000 nits | 18.5% |
| 222place_logo.jpeg | 400×400 | BT.2020 · PQ9 16 0 1 | 10,000 nits | 13.6% |
| knotapi_logo.jpeg | 400×400 | BT.2020 · PQ9 16 0 1 | 10,000 nits | 11.4% |
| programmers_force_logo.jpeg | 400×400 | BT.2020 · PQ9 16 0 1 | 7,699 nits | 8.7% |
Measured 29 August 2026.
How we verify
Every figure on this page comes from decoding the file, not from reading its description. The method is reproducible:
- Read the embedded ICC profile and extract its cicp tag, which is the reliable signal that a JPEG’s samples are PQ rather than sRGB.
- Decode the pixel data with colour management switched off, so the stored code values are seen rather than a tone-mapped rendering of them.
- Treat those bytes as PQ and convert each channel to absolute luminance through the SMPTE ST 2084 EOTF.
- Compute BT.2020 luminance per pixel, then take the peak and the share above the 203-nit reference.
The same code path runs in this site’s test suite, so the numbers regenerate rather than being transcribed by hand.
What we do not claim
These are files, measured. We are not asserting where any of them was published, when, or by whom — a filename is not a source. Brand attributions belong on this page only with a verifiable public URL and a capture date, and none is claimed here yet.
Try your own logo
The editor shows the same numbers for your file: encoded peak, glow coverage, the exact-colour ceiling for whatever colour you picked, and a verification pass over the finished bytes.
Make your logo glow →What are the numbers?
Every figure below is either a published constant or a value this tool computes. None is an estimate, and each one is traceable to the standard beside it.
| Measurement | Value | Source |
|---|---|---|
| Diffuse white reference | 203 nits | ITU-R BT.2408 |
| PQ maximum | 10,000 nits | SMPTE ST 2084 |
| 8-bit PQ code for diffuse white | 148 | Computed |
| 8-bit PQ code for 1,600 nits | 205 | Computed |
| Rec.2020 white ceiling | 10,000 nits | Computed |
| Rec.2020 green ceiling | 6,780 nits | Computed |
| Rec.2020 red ceiling | 2,627 nits | Computed |
| Rec.2020 blue ceiling | 593 nits | Computed |
| Blue's share of Rec.2020 luminance | about 6% | Computed |
| LinkedIn logo size | 400 × 400 px | Measured |
| LinkedIn post and ad size | 1200 × 627 px | Measured |
| LinkedIn JPEG re-encode | quality 90, 4:4:4 | Measured |
| Price | Free, no account | This site |
Where do these figures come from?
This tool implements published standards rather than an in-house approximation of them. Each specification below defines part of what the exported files contain.
- ITU-R BT.2100
Defines the HDR system this tool encodes into — BT.2020 primaries with the PQ transfer function.
- ITU-R BT.2408
Sets diffuse white at 203 nits, which is the reference every brightness figure here is measured against.
- SMPTE ST 2084
The PQ transfer function itself: absolute luminance from 0 to 10,000 cd/m², independent of display.
- ITU-T H.273
The code points behind cICP 9 16 0 1 — colour primaries, transfer characteristics, matrix and range.
- W3C PNG Third Edition
Specifies the PNG cICP chunk, which is how an HDR PNG declares itself without an ICC profile.