SMPTE ST 2094 SMPTE ST 2094: Dynamic Metadata for Color Volume Transform (DMCVT)

What SMPTE ST 2094 is

SMPTE ST 2094 standardizes dynamic metadata: per-scene or per-frame data that tells a display how to map a master graded on a bright reference monitor onto a dimmer, narrower-gamut consumer screen. Static HDR10 metadata (SMPTE ST 2086 plus MaxCLL and MaxFALL) describes the whole program once. Dynamic metadata is "specific to each frame or each scene of the video," in the words of the HDR video overview. SMPTE published ST 2094 in 2016 as six parts under the title Dynamic Metadata for Color Volume Transform.

The structure is a core plus applications. ST 2094-1 defines the common metadata items and structures. ST 2094-2 covers key-length-value encoding and the MXF container; DisplayDaily's account of Lars Borg's SMPTE webinar noted it was the one part not yet approved at the time. The four application parts are ST 2094-10 from Dolby (parametric tone mapping), ST 2094-20 from Philips (parameter-based color volume reconstruction), ST 2094-30 from Technicolor (reference-based color volume mapping) and ST 2094-40 from Samsung (scene-based color volume mapping). I did not read the standards text itself, which SMPTE sells, so this entry relies on DVB, SMPTE-presenter and encyclopedia descriptions.

How the application parts work

DVB's HDR dynamic mapping webinar gives the clearest engineering description. ST 2094-10 is "also known as Dolby Vision": the metadata describes content characteristics and drives a parametrically defined tone-mapping curve with three primary elements, deep shadow, mid-tone and highlight regions. ST 2094-40 is "also known as HDR10+": it uses a curve with a linear part up to a designed point where the image is left unchanged, then a shapeable Bezier curve above that point to adjust tones in any luminance range, plus percentile statistics on scene value distribution so the display can optimize on a frame basis. ST 2094-20 and -30 are the parameter and reference-based color volume methods that feed the SL-HDR family.

Several features are common. A single stream can hold multiple metadata tracks, each targeting a different display luminance such as 500 cd/m² or 1,000 cd/m², and a track specifies the time frame and the curve. Windowing lets different tone maps apply to different areas of one frame, though Borg knew of no implementation. Because the four methods are "quite different in their methodology," Borg said they cannot be transcoded into one another, so a producer picks one for mastering and distribution.

ST 2094-10 vs ST 2094-40

These two are the ones consumers meet. ST 2094-10 is Dolby's; Dolby Vision can represent up to 10,000 nits and requires mastering displays of at least 1,000 nits peak, per the HDR video overview, and the Dolby Vision entry covers the profile layer. ST 2094-40 is Samsung's, and HDR10+ is built on it: HDR10 base layer, 10 to 16 bits per channel, up to 10,000 nits, announced 20 April 2017. Licensing differs too, since HDR10+ Technologies describes its program as royalty-free with an annual administration fee for certain adopters, while Dolby licenses its format commercially. The HDR video overview puts HDR10+'s annual fee cap at $10,000. Both families list 10,000 nits as their peak, against roughly 100 nits for SDR.

The DVB document also frames the goal. Its commercial requirement is that dynamic mapping adapt the HDR signal to the more limited luminance and chromaticity capabilities of consumer displays compared with the grading reference monitor. SL-HDR2 carries five groups of information (HDR picture characteristics, SDR picture characteristics, luminance mapping variables, color correction variables and optional gamut mapping variables) in a parameter-based or table-based mode. The static baseline is ST 2086, the mastering display color volume, with MaxCLL and MaxFALL. PQ can encode up to 10,000 cd/m², but HDR10 content is commonly mastered at 1,000 to 4,000 nits, which is the gap dynamic metadata exists to manage.

Where ST 2094 lives

The metadata rides in the video bitstream. The HDR video overview says ST 2094 is carried in HEVC SEI messages, in ETSI TS 103 433 and in CTA-861-G, and ETSI TS 103 572 (October 2020) covers signaling and carriage of ST 2094-10. The DVB profile in ETSI TS 101 154 v2.6.1 (September 2019) lists three dynamic mapping solutions, ST 2094-10, ST 2094-40 and SL-HDR2, delivered as "user data T.35" SEI messages embedded in the elementary stream. Each is optional and backward compatible with PQ10 receivers, which can ignore the metadata, and multiple solutions can coexist in the same bitstream. DVB also noted that dynamic mapping generally helps low-end HDR displays more than high-end ones.

In products, ST 2094-10 appears as Dolby Vision on UHD Blu-ray and streaming, and ST 2094-40 appears as HDR10+ on Samsung, Panasonic and other sets plus Prime Video. The newer HDR10+ Advanced and Dolby Vision 2 extend those families. The HDR10+ press material states support for up to 10,000 nits at 10 to 16 bits per channel, and ITU-R BT.2100-3 (02/2025) is the current HDR television production recommendation that these formats sit beside.

Common confusions

ST 2094-10 is not the whole of Dolby Vision, since Dolby's metadata levels and profiles have evolved; I did not verify how today's Dolby Vision RPU maps to the 2016 text. SL-HDR2 is a separate Technicolor and Philips-lineage approach that DVB lists alongside the SMPTE parts. And "dynamic metadata" does not mean every frame is re-graded: it is a set of parameters, so the quality depends on how the display's tone mapper uses them. Read static vs dynamic HDR metadata for the end-user view.