Is Audio Description Legally Required? EU and UK Rules

There is no single European law that says every video must carry an audio description. That is the first thing to know, and it is why the question keeps getting answered badly. What exists instead are three separate regimes that overlap: a directive that governs how broadcasters and streamers must improve access over time, an accessibility act that governs the products and services people use to reach that content, and a technical standard that defines what counts as a description at all.

This page pulls the actual sources together in one place: which instrument imposes which duty, what the standards say a description has to do, what the quotas are where quotas exist, and what a broadcast-ready description track looks like once it reaches the studio.

Three regimes, three different duties

The Audiovisual Media Services Directive is where the content-side obligation lives. Article 7, as reinforced by Directive (EU) 2018/1808, requires member states to ensure that services under their jurisdiction are made continuously and progressively more accessible to people with disabilities through proportionate measures. Audio description is named as one of the means of achieving that, alongside subtitling for the deaf and hard of hearing, spoken subtitles and sign language. Note the wording: continuous and progressive improvement, not a fixed percentage.

The European Accessibility Act, Directive (EU) 2019/882, has applied since 28 June 2025. It is often reported as the law that made audio description mandatory. It is not, or at least not directly. Its audiovisual provisions cover access to the content: the player, the electronic programme guide, the app, the set-top box and the mechanisms that let a viewer find an accessible version and switch to it. If a service offers an audio description track, the Act requires that the track be transmittable, findable and selectable with assistive technology. The Act contains transitional arrangements in Article 32 for services already running before that date.

The third layer is technical. EN 301 549 and the WCAG success criteria define what a compliant description actually is, and they are what an auditor tests against. National laws point at these standards rather than restating them.

What the standards say a description must do

WCAG splits the obligation across two levels. Success criterion 1.2.3, Audio Description or Media Alternative (Prerecorded), sits at Level A and lets an author choose: either an audio description, or a full text alternative for the video. Success criterion 1.2.5, Audio Description (Prerecorded), sits at Level AA and removes the choice. At AA, the description track is required. Extended audio description, where the video pauses to make room for longer description, is criterion 1.2.7 at Level AAA and is rarely mandated.

That matters because AA is the level most European public-sector and accessibility legislation adopts as its benchmark. The practical effect is that an organisation which believed a transcript would cover it at Level A finds it does not at Level AA.

EN 301 549 V3.2.1, published in March 2021, handles the other half. Clause 7 covers ICT with video capabilities and applies to players, media software, smart TVs and set-top boxes rather than to the content. It requires that a product displaying video with synchronised audio provide a mechanism to select and play an available audio description, and that users be able to control the level of the description separately from the main programme audio. Content compliance and product compliance are separate tests; both have to pass.

For the writing itself, the reference document is ISO/IEC TS 20071-21:2015, which gives guidance to description scriptwriters, narrators and the organisations delivering description across recorded video, broadcast, cinema, live drama and museum settings. It is a technical specification rather than a certifiable standard, and it deliberately addresses the subjective judgement involved rather than pretending description is mechanical.

Quotas: where they exist and what they are

The AVMSD does not set numbers. Some member states and the UK do.

The clearest quota regime is the UK's. Under sections 303 to 305 of the Communications Act 2003, Ofcom sets targets in its Code on Television Access Services, last updated on 30 January 2017. For channels above the audience-share threshold the audio description target rises from 2 per cent of non-excluded programming in the first year to 10 per cent from the fifth anniversary onwards, where it stays. The corresponding subtitling target reaches 80 per cent and signing 5 per cent. Channels with a UK audience share of 0.05 per cent or less are excluded entirely.

The Code also carves out an honest exception. Ofcom accepts that description may be impracticable for music and news programming, where there is little space in the dialogue to insert it and less need for it, but requires broadcasters relying on that exception to train producers, editors and presenters in describing the significance of images on air, and to file a statement of that training within twelve months.

Germany took the opposite route. The Medienstaatsvertrag obliges broadcasters and on-demand providers to report on the state of accessibility in their offering every three years and to improve it, but sets no percentage. The German Federal Agency for Accessibility describes the reporting duty as the implementation of Article 7(3) of the AVMSD. Advocacy groups have argued for years that reporting without a target changes little: an evaluation by the TV für Alle project, reported by Die Neue Norm in September 2020, put the share of German television programming carrying audio description at roughly 4 per cent, against roughly a third for subtitling.

Germany's BFSG and what it actually reaches

Germany implemented the European Accessibility Act through the Barrierefreiheitsstärkungsgesetz (BFSG). Section 3 BFSG states the principle: products placed on the market and services offered must be accessible, meaning findable, reachable and usable by people with disabilities in the customary way, without particular difficulty and in principle without outside help. The concrete requirements sit in the implementing regulation, the BFSGV.

Two limits are worth knowing before anyone panics about a corporate video library. First, section 3(3) BFSG exempts micro-enterprises that provide services, defined as fewer than ten employees and annual turnover or balance sheet total up to two million euros. Second, the BFSGV's audiovisual provisions target consumer terminal equipment used to access audiovisual media services and services in electronic commerce, not every video a company publishes.

Where a video sits inside a service that does fall in scope, the WCAG Level AA benchmark pulls audio description back in through the front door. That is the route by which most in-house teams end up needing a description track: not because a law named their video, but because the service it lives in has to meet AA.

What a broadcast-ready description track looks like

Once the legal question is settled, description becomes a production problem, and here the most detailed public specification is not a law but a platform document. Netflix publishes its Audio Description Style Guide v2.5, last revised in April 2023. It is worth reading even if you will never deliver to Netflix, because it is the closest thing the industry has to a written craft standard.

On casting, it asks that the describer's voice be chosen to complement or contrast deliberately with the dominant voices in the programme, that age and accent match the content and its audience, and that the same voice talent carry across all episodes and seasons of a series and across sequels where possible. It states plainly that a sense of the content matters more than vocal characteristics.

On delivery, the guide asks for a present-tense, third-person, conversational read that reflects the pace of the scene: flowing and unhurried in a romantic sequence, quicker and more staccato in a fight or chase. Description over dialogue is a last resort. Music, sound effects and intentional silence are interrupted only for information that cannot wait.

On the mix, the numbers are specific:

• For a 5.1 printmaster, dip the centre channel for descriptive events; left and right generally no more than -6 dB, and sparingly to -12 dB only where unavoidable.
• Dip the original mix by 6 to 12 dB at the mixer's discretion, so the description is intelligible with the natural presence of the original dialogue underneath.
• Side-chained compressors should use an attack time no shorter than 2 ms and no longer than 15 ms.
• Transitions into and out of a dip should take no more than 5 seconds, with no abrupt level changes.
• Mix level follows the platform's loudness and true-peak specification; the description is never raised above it to compete with a loud passage.
• Avoid noise reduction. A clean recording should need almost no EQ or compression.

Ofcom's own guidance in Annex 4 of the Access Services Code lands in the same place from a different direction: delivery should be steady, unobtrusive and impersonal but not monotonous, present tense, and the describer should avoid the phrase “we see” and film-craft terms like camera angles. Both documents insist on describer consistency across a series, for the same reason: to a regular viewer, the describer is part of the programme.

If you are also mixing for a platform delivery spec, our guide to voice-over loudness standards and LUFS targets covers the numbers the description track has to land inside.

Disclosure if the description is synthetic

One newer wrinkle: if the description is generated or manipulated by AI rather than voiced by a person, a separate transparency duty applies in the EU. Article 50 of the AI Act has applied since 2 August 2026 and reaches synthetic and cloned audio. We covered the scope in detail in what the EU AI Act requires voice work to disclose. Accessibility compliance and AI disclosure are separate obligations, and meeting one does not discharge the other.

Where to start if you need a description track

Work in this order. Establish which regime actually catches you: a broadcaster or on-demand service falls under the AVMSD transposition in its member state, a consumer-facing digital service falls under the EAA and its national implementation, and almost everyone else arrives at description through a WCAG Level AA commitment rather than a statute.

Then treat the script as the deliverable it is. A description script is written work with its own constraints: it has to fit the gaps, prioritise what is plot-critical, and stay consistent across a series. Budget for it separately from the recording.

Then cast for the content rather than for a generic “neutral narrator”. The standards agree on this point more than on anything else. And cast in the language of the programme: the description track has to match the language of the audio it sits in, which for multi-market releases means a native describer per language. Voicfy works with native-language voice talent across English, German, Spanish, French, Dutch, Italian and Portuguese, and you can brief a description project and receive quotes from describers who have worked to these specifications.