HDR metadata is the descriptive data carried alongside high dynamic range video that tells a display how the content was mastered: which mastering monitor was used, how bright the brightest pixels are and, in dynamic formats, how each scene or frame should be tone mapped. In HDR video, it is the link between the colorist’s grading room and the very different screens on which viewers watch.
Metadata matters because consumer displays vary widely in peak luminance and color volume. A film graded on a 1,000-nit or 4,000-nit monitor must be adapted to a television that may reach only 500 nits. Without reliable metadata, the display must guess, which leads to clipped highlights, crushed detail or an image that looks too dark.
Static and dynamic HDR metadata
HDR metadata falls into two families. Static metadata describes the whole program with one set of values, while dynamic metadata provides new values per scene or even per frame. The family determines how precisely a display can adapt the picture.
| Property | Static metadata | Dynamic metadata |
|---|---|---|
| Granularity | One set for the whole program | Per scene or per frame |
| Main standards | SMPTE ST 2086, CTA-861.3 | SMPTE ST 2094 family |
| Formats using it | HDR10 | HDR10+, Dolby Vision |
| Tone mapping quality | Compromise across all scenes | Optimized for each scene |
| Production effort | Low, values measured once | Analysis or trim passes required |
Static metadata forces the display to choose one tone mapping curve for the entire film. A program that contains both a dim night scene and a bright desert scene receives the same treatment, so either the night scene is darkened unnecessarily or the desert highlights clip. Dynamic metadata solves this by describing the luminance distribution of each scene, allowing the display to keep shadows open in dark scenes and compress highlights only where needed.
The fields of static HDR10 metadata
HDR10 static metadata combines the mastering display color volume defined in SMPTE ST 2086 with two content light level values defined in CTA-861.3. Together they describe both the monitor the colorist used and the brightness actually present in the content.
| Field | Meaning | Typical values |
|---|---|---|
| Display primaries | Chromaticities of red, green, blue on the mastering monitor | P3-D65 or Rec. 2020 |
| White point | Chromaticity of mastering white | D65 |
| Max mastering luminance | Peak brightness of the mastering monitor | 1,000 or 4,000 nits |
| Min mastering luminance | Black level of the mastering monitor | 0.0001 to 0.005 nits |
| MaxCLL | Brightest single pixel in the program | Often 400 to 4,000 nits |
| MaxFALL | Highest frame-average light level | Often 100 to 400 nits |
MaxCLL and MaxFALL are measured from the finished content, not from the monitor. They tell the display how much headroom it really needs. If MaxCLL is 800 nits, a 1,000-nit television can show the program almost without tone mapping, even though the mastering display reached 4,000 nits. Values of zero mean unknown, and displays then fall back to their own assumptions.
Dynamic metadata in HDR10+ and Dolby Vision
Dynamic metadata standards are grouped under SMPTE ST 2094, with separate applications for different systems. HDR10+ uses the application defined in ST 2094-40, and Dolby Vision uses its own metadata with a mapping related to ST 2094-10.
In HDR10+, an analysis tool measures each scene and produces luminance percentiles and Bezier tone mapping curve parameters for target displays. Colorists can refine the results, but much of the process can be automated. In Dolby Vision, an analysis pass produces per-shot values, and the colorist creates trim passes for target displays such as 100-nit SDR and 600-nit or 1,000-nit HDR. The display then interpolates between those trims for its own capability.
HLG takes a different route. Hybrid log-gamma is a scene-referred system designed for broadcast and requires no mandatory metadata, because the signal itself is meant to display acceptably on both SDR and HDR screens. This simplicity is why HLG is common in live television.
How metadata travels from master to screen
Metadata must survive every step of the delivery chain: mastering, encoding, packaging, transmission and the connection to the display. A single step that drops or rewrites it can break the intended presentation.
| Stage | Carrier | Notes |
|---|---|---|
| Mastering | Sidecar XML or embedded in master files | Dolby Vision XML, IMF packages |
| HEVC and VVC video | SEI messages in the bitstream | Mastering display and light level SEI |
| AV1 video | Metadata OBUs | Equivalent HDR fields |
| Containers | MP4 boxes, Matroska elements | Duplicate of the stream values |
| HDMI link | InfoFrames | Sent from player to display |
Problems often occur at transcoding and playback. An encoder that does not pass through SEI messages produces a file with PQ signal but no metadata, and a streaming device may send generic values over HDMI instead of the real ones.
Practical guidance and common pitfalls
Correct HDR metadata is a quality-control item, not an afterthought. Delivery specifications for streaming platforms usually require exact mastering values and measured light levels, and files are rejected when they are missing or implausible.
- Copy mastering values from the actual monitor used for grading, not from a generic template.
- Measure MaxCLL and MaxFALL on the final render, after all titles and effects are applied.
- Watch for isolated bright pixels, such as specular highlights or graphics, which inflate MaxCLL and can make displays tone map too aggressively.
- Verify metadata after every transcode with an analysis tool that reads SEI messages or container fields.
- Keep SDR versions consistent, because HDR-to-SDR conversion in Dolby Vision depends on the trim metadata.
Metadata cannot fix a poor grade. It only describes the content, so an image graded beyond the realistic capability of consumer displays will still lose detail after tone mapping, even with perfect dynamic metadata.
The idea in brief
HDR video can contain very bright and very dark parts. Every television, however, can show only a certain range of brightness. HDR metadata is a small set of instructions sent with the video. It says how bright the film was made to look and how bright its brightest moments really are.
The television reads these instructions. It then decides how to fit the picture into what it can show. Basic metadata gives one instruction for the whole film. Advanced formats give new instructions for each scene, which usually looks better.
A practical scenario
A post-production house delivers a 90-minute HDR feature to a streaming platform. The film was graded on a 1,000-nit reference monitor with P3-D65 primaries inside a Rec. 2020 container, and a Dolby Vision analysis with trims for 100 nits and 600 nits was completed.
During quality control, the team measures MaxCLL of 1,450 nits, caused by a few frames of an explosion graphic that exceed the monitor’s peak. The colorist limits the graphic to 1,000 nits, the file is re-rendered and measured again at MaxCLL 1,000 and MaxFALL 260. The HEVC deliverable passes validation with correct SEI metadata, and the film tone maps cleanly on both premium and midrange televisions.