The short answer
AI audio description (in German, KI-Audiodeskription) means using artificial intelligence for parts of the audio description (AD) process: mainly narrating descriptions with synthetic voices, plus detecting scenes and gaps in dialogue and, increasingly, drafting first versions of the text. German public broadcasters already use it in their streaming offers: ZDF, WDR, NDR, MDR and Radio Bremen all narrate some audio description with synthetic or AI voices, mostly for documentaries. The descriptions themselves are still written and signed off by trained editors.
The legal picture is also more specific than many guides suggest. Germany's accessibility law, the Barrierefreiheitsstärkungsgesetz (BFSG), mainly governs whether players and services make audio description technically accessible. The obligation to expand how much content has AD comes from the Medienstaatsvertrag (Interstate Media Treaty).
What is AI audio description?
Audio description is an additional audio track that tells blind and partially sighted viewers what is happening on screen during pauses in the dialogue: who is acting, where the scene takes place, what happens. Traditionally it is produced in three steps: a describer analyzes the film and writes the script, an editor approves it, and a narrator records it in a studio. AI supports these steps to very different degrees:
- Synthetic voices: mature and already on air. They replace the studio session and make corrections possible without booking a narrator again.
- Scene and dialogue detection: reliable. Software finds cuts and pauses where descriptions fit, instead of someone searching for every gap by ear.
- Automatic script drafts: still early. Generated descriptions tend to over-describe, miss what matters to the plot, or get the tone wrong, and need a full human edit.
The realistic state in 2026: people write and take responsibility, AI speeds up the voice and the mechanics. ZDF's audio description requirements state that every description should be approved by trained editors, and list synthetic and AI voices as an option alongside human narrators.
What the law requires from streaming platforms in Germany
BFSG and the European Accessibility Act: access
The BFSG implements the European Accessibility Act (Directive (EU) 2019/882) and has applied since June 28, 2025. For services that provide access to audiovisual media, the EAA essentially requires that existing access services such as subtitles, audio description and sign language are transmitted completely, in adequate quality and in sync; that viewers can control them; and that programme guides show whether they are available. General requirements for accessible websites and apps apply on top. The BFSG does not set a quota for how much content must carry AD.
Medienstaatsvertrag: expanding the content
The content obligation sits in Section 7 of the Medienstaatsvertrag: broadcasters should add accessible offers within their technical and financial means and expand them steadily and step by step. Section 76 applies the same rule to TV-like on-demand services (fernsehähnliche Telemedien), which covers streaming platforms; providers with low turnover or small audiences can be exempt. Nationwide private broadcasters report to their state media authority (Landesmedienanstalt), and ARD, ZDF and Deutschlandradio to their supervisory boards, at least every three years in action plans (Aktionspläne) covering measures and progress.
In practice, a platform that has to show a growing share of described content needs a production chain that scales with its catalogue. That is where AI comes in. This overview is not legal advice, and services aimed at other markets, such as the US, are governed by different rules.
How German-language broadcasters use AI today
ZDF reports that 32 percent of its content carried audio description in 2025, and the share of described fiction rose from 61.9 percent in 2023 to 79.8 percent in 2025. The documentary series "Terra X" has used an AI-optimized synthetic voice for its AD since April 2024, and under its "Access 2028" plan ZDF will extend AI-assisted AD to further documentaries by 2028. According to reporting by the ver.di union magazine, MDR, WDR, NDR and Radio Bremen also use synthetic or AI voices, with WDR using them regularly for documentaries.
In Austria, ORF's 2025–2028 accessibility action plan reports audio description on about 8.8 percent of its total 2025 output and on more than 84 percent of suitable prime-time programming on ORF 1 and ORF 2, with synthetic AD in growing use. The pattern is consistent: AI voices arrive first in genres with a factual narrative style, while high-end fiction keeps human narrators.
Quality and acceptance: what audiences expect
The German Federation of the Blind and Partially Sighted (DBSV) does not reject AI voices in principle, but considers professional human voices necessary for high-quality audio description. Three rules have emerged across the industry: AI use is disclosed, as the AI principles of ARD and ZDF require; every description is editorially approved; and synthetic voices are chosen by genre rather than applied across the board. For the text itself, the shared principles for audio description in German-speaking countries, which ZDF's requirements build on, are the reference.
Audio description and localization: multilingual catalogues
For platforms with catalogues in several languages, AI changes localization most. An approved AD script can be translated, adapted and voiced with a synthetic voice in the target language without organizing a studio production for each language. It helps to separate the services that are often bundled as "media accessibility and localization":
- Subtitles translate dialogue for viewers who speak another language.
- Closed captions and SDH add sound effects, music and speaker identification for deaf and hard-of-hearing viewers.
- Audio description describes the picture for blind and partially sighted viewers.
- Dubbing replaces the dialogue track in another language.
These services share technology such as speech recognition and speech synthesis, but they need different expertise. If you buy them bundled from one provider, evaluate the quality of each service separately.
What streaming platforms should check in AI audio description
- Editorial control: can every description be reviewed, changed and approved before release?
- Voices and languages: are natural-sounding synthetic voices available in every language you need, including German, and can pace and level be matched to the film's mix?
- Timing: are overlaps with dialogue and important sounds reliably detected?
- Delivery formats: does the workflow produce a separate AD audio track that can be flagged as audio description in HLS and DASH streams, so players and programme guides recognize it as the EAA expects?
- Content security: unreleased masters are valuable. Must material be uploaded to a cloud service, or can it stay inside your own environment?
- Documentation: can you record which titles were described, when and how? That makes Medienstaatsvertrag action plans much easier to write.
Where Synchrogen fits
Synchrogen is a desktop audio description editor for Windows and macOS. It detects scenes and dialogue, shows the gaps on a timeline, and lets describers write and time descriptions in a timeline or script view. Local text-to-speech, including German voices, narrates the script inside the app, and exports include an AD audio track and a WebVTT file. Video is processed on the describer's own machine rather than uploaded to a cloud service, which matters for unreleased productions. Every description is reviewed and approved by a person. Synchrogen is in closed beta; AD service providers and streaming platforms can get in touch.
FAQ
Is audio description mandatory for streaming platforms in Germany?
There is no fixed quota. Under Sections 7 and 76 of the Medienstaatsvertrag, broadcasters and TV-like on-demand services should expand accessible offers within their means and report on progress. The BFSG additionally requires that existing audio description is delivered completely, in sync and under the viewer's control in the player.
Can synthetic voices be used for audio description?
Yes. ZDF, WDR, NDR, MDR and Radio Bremen already use them, mostly for documentaries. Standard practice is to disclose AI use and to approve every description editorially; for high-end fiction, organizations of blind people still prefer human narrators.
Can AI create audio description fully automatically?
Not at broadcast quality. Voice, scene detection and timing can be largely automated. Deciding what matters to the story, and approving the result, remain human work.
What is the difference between subtitles, closed captions and audio description?
Subtitles render dialogue as text, closed captions and SDH add sound effects and music for deaf and hard-of-hearing viewers, and audio description narrates the picture as an extra audio track for blind and partially sighted viewers.