SEI

Supplemental Enhancement Information

Services →
Introduced in Rel-8

SEI is a metadata message format used in video coding standards to convey auxiliary, non-essential information that enhances the decoding, display, or interpretation of the video bitstream.

Category
Services
Introduced
Rel-8
Where
Services › Codecs
Specifications
15 specs
SEI Description Purpose Related Specifications

Description

Supplemental Enhancement Information (SEI) is a standardized mechanism for embedding auxiliary metadata within a video bitstream, as defined by video coding standards adopted by 3GPP for multimedia services. It is part of the syntax of codecs like H.264/AVC (Advanced Video Coding) and H.265/HEVC (High Efficiency Video Coding). SEI messages are contained in the Network Abstraction Layer (NAL) units of the bitstream. They carry information that is not required for the normative decoding process to produce correct sample values but is beneficial for improving the user experience, guiding display processes, enabling advanced features, or providing debugging information.

SEI messages work by being inserted at the encoder and transported alongside the coded video slices. The decoder parses these messages and can use the information they contain. The syntax and semantics of SEI messages are rigorously defined. Each SEI message has a specific payload type and payload structure. Common examples include "pic_timing" which provides clock timestamp information for synchronizing decoding and display, "user_data_registered_itu_t_t35" for carrying vendor-specific or application-specific data, and "recovery_point" which signals how to recover from errors or seek points. The messages are unidirectional (from encoder to decoder) and do not require acknowledgment.

From a 3GPP network perspective, SEI messages are treated as part of the video user plane. During multimedia session establishment via protocols like SIP and SDP, the support for certain SEI messages can be negotiated. The core network and radio access network transparently transport the bitstream containing SEI. However, media-aware network elements, like a Media Resource Function (MRF) or application servers, might parse and even generate or modify SEI messages to adapt content for different devices or network conditions. For example, an MRF might insert frame packing arrangement SEI messages for stereoscopic 3D video services.

The role of SEI is crucial for enabling enhanced video services over mobile networks. It supports features like dynamic adaptation, trick modes (fast forward, rewind), color volume transformation (HDR), and content description. By standardizing this metadata carriage, SEI ensures interoperability between different vendors' encoders, decoders, and middleware, allowing for a rich ecosystem of video applications within the 3GPP multimedia framework.

Purpose & Motivation

SEI exists to solve the problem of conveying non-essential but highly valuable control and descriptive information alongside the compressed video essence, without altering the core decoding standard. Before its adoption, such auxiliary data had to be sent out-of-band (e.g., in a separate RTP header extension or a completely different channel), complicating synchronization and increasing system complexity. SEI provides an in-band, standardized carriage mechanism.

The primary motivation was to enable advanced video features and improve robustness and usability. For instance, precise display timing information (pic_timing SEI) is critical for smooth playback and lip-sync in multimedia streaming. Pan-scan rectangles allow for adjusting the viewing area on different display aspect ratios. SEI messages for buffering period and recovery points are vital for adaptive streaming protocols like DASH and HLS, which are widely used in 3GPP's MBMS and streaming services. They help decoders manage buffers and recover from packet loss or seek operations efficiently.

Historically, SEI was incorporated from the ITU-T and ISO/IEC MPEG video standards into the 3GPP multimedia specifications to ensure mobile devices could interoperate with broad video ecosystems. Its creation addresses the limitations of having a video bitstream that only contains pixel data, by adding a layer of 'intelligence' that guides how the video should be handled post-decoding. This allows service providers to offer a consistent, high-quality video experience across diverse networks and devices, which is a fundamental goal of 3GPP's Packet-Switched Streaming Service (PSS) and Multimedia Broadcast/Multicast Service (MBMS).

Evolution Across Releases

Rel-8 Initial

Introduced with the adoption of H.264/AVC as a primary video codec for 3GPP multimedia services. SEI messages from the H.264 standard, such as pic_timing and user_data_unregistered, became part of the video bitstream syntax for services like PSS and MBMS, enabling basic enhancement information transport.

Explore further

Broader topics and technologies where SEI plays a role.

Defining Specifications

3GPP specifications that define or reference SEI, with the latest known release. Sourced from the 3GPP document catalog — see methodology.

SpecificationTitleRelease
TS 26.114 vj10 IMS Multimedia Telephony Media Handling Rel-19
TS 26.116 vj00 TV Video Formats for 3GPP Services Rel-19
TS 26.118 vj00 Virtual Reality Media Formats Rel-19
TS 26.223 vj00 IMS Telepresence Client Specification Rel-19
TS 26.346 vj20 MBMS User Services Media Codecs & Protocols Rel-19
TS 26.804 vj10 5G Media Streaming Extensions Study Rel-19
TS 26.855 vj00 Study on Film Grain Synthesis Rel-19
TR 26.862 vh00 Immersive Teleconferencing & Telepresence for Remote Terminals Rel-17
TR 26.906 vj00 HEVC Evaluation for 3GPP Services Rel-19
TR 26.919 vj00 Study on 5G Conversational Media Handling Rel-19
TR 26.946 vj00 MBMS User Services Overview Rel-19
TR 26.948 vj00 Video enhancements for 3GPP Multimedia Services Rel-19
TR 26.949 vj00 TV Service Profiles for 3GPP Networks Rel-19
TR 26.955 vj00 Video Codec Analysis for 5G Services Rel-19
TR 26.962 vj00 ITT4RT Operation and Usage Guidelines Rel-19
Patrick Zandl

About the author: Patrick Zandl (b. 1974)

Telecommunications specialist, technology journalist (founder of the Mobil server), and developer who has been running since 2025 — the largest Czech-language resource on AI-assisted programming. Formerly Chief Wizard Architect at Prusa3D and head of development for Turris at CZ.NIC; currently a consultant and instructor on AI implementation in companies.