Tabitha Allum of Built for Good explained how WatchWord frees the closed captions “trappedˮ on the Digital Cinema Package. Steve Case of Omnex Pro Film walked through the journey of hearing-impaired and audio description tracks from distributor to seat. Mereijn van der Heijden showed how Earcatch made every Dutch cinema accessible without installing a single box. Irene Muñoz López of the University of Surrey surveyed where research into artificial intelligence is heading. A panel chaired by Weronika Kolodziej of Deluxe asked how accessibility can become an expectation rather than a special arrangement.
Swap “Digital Cinema Packageˮ for “Interoperable Master Format packageˮ, “projection boxˮ for “playout galleryˮ and “front-of-houseˮ for “programme guide and appˮ, and most of it could have been a broadcast conference. What follows is our attempt to join those dots, and to be honest about where a managed playout and media services partner like Levira can help.
Cinemaʼs distribution chain is a useful mirror for televisionʼs. Facilities such as Deluxe and MPS master a Digital Cinema Package. It travels by satellite, broadband or courier to the cinemaʼs storage, where a composition playlist selects the picture, sound and text to be played together. Within it, the hearing-impaired and audio description tracks sit as mono audio on channels 7 and 8, following the recommendations of the Inter-Society Digital Cinema Forum, while captions travel as separate timed-text data. As Steve Case put it, the delivery system “transfers the package. It does not create the access tracks.ˮ Having a hearing-impaired track does not guarantee that you have captions.
Television has exactly the same shape. The eveningʼs closing quiz made the link explicit: the Interoperable Master Format, defined in SMPTE ST 2067, shares many of its concepts, including the composition playlist, with the cinema package. Subtitle files, audio description mixes and signed versions arrive as components of a master or a delivery package. They then have to survive ingest, transcoding, scheduling, frame-rate conversion, playout, encoding and every downstream platformʼs packaging before they reach a viewer. Each of those steps is a chance for an access component to be dropped, mislabelled or left out of sync, and nobody notices until a viewer complains.
That is the real lesson. Access services rarely fail inside one system. They fail at the joins between suppliers, where each party assumes someone else is checking. It is a familiar problem for Levira, because holding those joins together is what a managed media services partner does.
Caseʼs favourite tool for avoiding the problem in cinema was simply reading the label. His illustrative playlist name, ExampleFilm_FTR1_F_ENENCCAP_GB12A_51HI
VI_2K_ST_20261002_FAC_SMPTE_OV, tells an operator at a glance whether hearing-impaired, visually impaired and closed caption components are present before the doors open. In playout, the equivalent is metadata that automation actually checks. If a schedule item promises audio description and subtitles, the system should confirm that they exist, are in the right language and are the right duration before the item goes to air, rather than an operator discovering the gap during transmission. We think access-service validation deserves to be a first-class quality control check, in the same way that loudness and aspect ratio already are.
The most repeated complaint of the evening was not about glasses or receivers. It was about adverts and trailers. Allum described patrons watching captions work perfectly on the foyer loop, sitting through a run of uncaptioned adverts in the auditorium, and concluding that the glasses were broken. Paul Willmott of Saffron Screen called it his pet hate, and asked for subtitled adverts and trailers so that staff can confirm a patronʼs glasses work before the feature starts. Most major releases now arrive with closed captions, a change Allum said has happened in just the last two years. The material around them has not caught up.
Anyone who runs a channel will recognise this. On linear television, promotions, interstitials and commercials often lack subtitles that the programmes either side of them carry. On free ad-supported streaming channels the problem is sharper, because adverts are frequently stitched in on the server, from a different supply chain, at the last moment. A viewer who relies on subtitles experiences the channel as a sequence of gaps, and quite reasonably concludes that the service is unreliable.
The fix is partly commercial, as it is for Built for Good in cinema, because advertisers need to be persuaded. It is also operational. A playout or channel assembly service can flag, and in time automatically remedy, any break material that arrives without the access components the surrounding schedule promises. Access continuity across breaks is a channel quality measure, and it belongs to whoever is accountable for the channel on air, not to the advertiserʼs afterthoughts.
If one idea ran through every cinema session, it was that the acceptance test is the customerʼs experience. Omnexʼs commissioning checklist ends at the customerʼs seat: the right content and routing, coverage at the front, back and edges of the auditorium, batteries that last the length of the film, clean devices, and staff who know how it all works. Case told of a multiplex manager handing over a receiver with an apology for having missed the training day. A questioner from a community of visually impaired athletes reported that every member of the group who still goes to the cinema has been let down by the technology. “How do we prove the service still works six months later?ˮ asked one Omnex slide. Television should ask itself the same question.
In broadcast, our rack is the playout chain and our seat is the viewerʼs television, phone or set-top box. Knowing that an audio description track is present at the output of the playout server is not the same as knowing that it is selectable, labelled and in sync on a smart television application, an operatorʼs set-top box or a streaming platform. Levira already provides round the clock operations for the channels we look after. The opportunity is to extend monitoring further downstream, to confirm that subtitles and audio description are reaching platforms in a usable form, and to report on it in a way that a channel owner or regulator would recognise. Accessibility that is checked at the sofa, and not only in the rack, is a reasonable thing for a managed service to promise.
Earcatch offered the eveningʼs most direct crossover. Because it serves both cinemas and Dutch television, it had to solve a problem that every broadcaster with a film catalogue knows well: films are made at 24 frames per second, while European broadcast runs at 25. Earcatchʼs software recognises which version is playing and plays the audio description back at the matching speed. A titleʼs accessibility work therefore stays usable as the film moves from cinema to broadcast and streaming, instead of being remade for each window.
In a traditional television workflow, a speed-changed film often means re-conforming or re-timing subtitle and audio description files by hand, or simply not carrying them over. As more cinema content arrives with closed captions and audio description as standard (in the Netherlands, almost every film financed by the Netherlands Film Fund now must have them), the downstream value of those assets is substantial, provided someone carries them across. Media processing that converts picture, sound, subtitles and audio description together, as one package, is quietly valuable work. It turns access services from a cost incurred for each window into an asset that follows the content.
Earcatchʼs other lesson is about where the access service lives. Before 2007 the Netherlands had no audio description at all, and a handful of Dolby ScreenTalk units installed that year did little for people who travelled across the country to find the devices uncharged. Earcatch put the service on the audienceʼs phone instead, synchronising by audio fingerprinting. Because the Netherlands still has no broadcast infrastructure for delivering audio description to the home, the same application now carries audio description for Dutch television too. It has around 35,000 users in the Netherlands and holds a Trusted Partner Network Blue Shield, demonstrating its commitment to content security. It protects everything, including subtitles, with digital rights management and, crucially, never releases content before transmission or release.
For channel operators, particularly smaller linear and free ad-supported services where adding and signalling an extra audio track on every platform is costly or simply not supported, a synchronised companion route is worth taking seriously. It does not replace in-stream access services, and should not. But a playout provider is unusually well placed to support it, because the playout chain already knows exactly what is going to air and when. Generating reference fingerprints, timing and access files as a by-product of playout, rather than as a separate post-production job, is the sort of joined-up working a managed service could offer without asking channel owners to change how they distribute.
Live content came up during the panel questions, and the answers carried straight across to live and near-live television. For theatre, Allum explained, Built for Good deliberately does not use live transcription, because it is not precise enough for a playwrightʼs words and struggles with character names, sound effects and descriptions of music that convey artistic intent. Instead its system matches pre-prepared captions to what is happening on stage in real time, using audio and cues from the lighting and sound desks. For live event cinema with a script, she suggested, that is the better route. Earcatch, for its part, delivered live audio description and captions for the 2021 Eurovision Song Contest to both the arena audience and the broadcaster, and the recordings are still available.
elevision already mixes these approaches: scripted segments subtitled from prepared text, and unscripted segments by respeaking or, increasingly, automatic speech recognition. The cinema and theatre experience is a reminder that the quality bar is set by the audience, not the technology, and that the cue data in a playout chain is an underused asset for timing prepared access content accurately. The SMPTE UK evening itself, live captioned through a cloud service fed from the sound desk, showed both the promise and the fragility (and the danger of corporate laptops with screensavers). Live channel work, including Leviraʼs role as an industry partner to the Media Talent Manifestoʼs On Air student broadcast, which goes live on 22 October, is exactly the kind of environment in which prepared and live access services can be shown working together in practice.
Irene Muñoz López made a point that television people should hear clearly. Cinema is a shared space and a social contract in which everyone compromises a little. At home, content can be adapted entirely to the individual. Her examples ranged from dialogue boost and picture-in-picture British Sign Language interpreters to personalised colour profiles for different forms of colour blindness, lighting control for people with epilepsy, and colour and light used to guide the attention of viewers who are easily distracted. She also relayed what deaf and hard of hearing people told an RNID project: something as simple as different colours for different speakers makes captions far easier to follow. United Kingdom television subtitling has long used colour to distinguish speakers, which suggests that on this point broadcast is ahead of cinema rather than behind it.
She was equally clear about the limits of artificial intelligence. Vision-language models can describe a scene, but they struggle with long-form content: they forget characters and what they have already said, and they hallucinate. Machine-generated audio description can sound robotic and out of tune with the scene, and machine-generated signing still lacks the expressiveness of a human interpreter, whom deaf audiences have no wish to lose.
That matches how we think about the technology at Levira. Artificial intelligence earns its place when it supports people by handling repetitive, data-driven work, and experienced operators remain accountable for the result. For access services, that means workflows in which the machine drafts and a person checks, adapts and signs off. It also means that the different versions a viewer might choose (standard, audio described, signed, dialogue enhanced) are produced, packaged and delivered reliably to every platform able to offer them. Streaming and free ad-supported services, which can carry multiple variants far more cheaply than a terrestrial multiplex, are where that personalisation will show up first.
The panel opened with an awkward fact. David Hancock, Chief Analyst at Omdia, found very little data on the use of accessibility in United Kingdom cinemas. What he could piece together suggested that only 2 to 3 per cent of screenings are subtitled, rarely at prime time, and that there are no accessible screenings of opening films at weekends. This is in a country where around 18 per cent of people are disabled, with a spending power Omdia estimates at £274 billion a year.
Where the numbers do exist, they change minds. Use of ODEONʼs WatchWord tripled within months to 1,400 uses across 15 sites in August, and Allum said the system there had already paid for itself. Earcatchʼs download figures, in the thousands for popular titles and the hundreds for art-house films, are what persuaded Dutch filmmakers and cinemas that it was worth supporting.
The audience is also wider than the label suggests. Built for Good finds that a quarter of caption users at The Hunger Games in London identify as neurodivergent, and that the biggest single group of caption users at the British Film Institute is Chinese students. Younger viewers increasingly watch with subtitles on by choice.
Television and streaming have an advantage that cinema mostly lacks: they can measure. Every streaming and free ad-supported platform knows when a subtitle or audio track is selected. Bringing that data together, per channel and per programme, and presenting it alongside the usual audience reporting would give channel owners the business case that cinema exhibitors are still trying to make. It is a modest piece of engineering for a partner that already operates streaming and distribution services, and it would turn accessibility from a compliance cost into a measurable audience.
Van der Heijdenʼs frustration was familiar from the delivery side of television. Earcatch can satisfy one studioʼs content security requirements completely, only for the next studio to arrive with a different list. Common guidelines that worked for all of them would help everyone, audiences included. Television faces the same fragmentation across broadcastersʼ and platformsʼ delivery specifications, each with its own expectations for subtitle formats, audio description mixes and metadata.
Standardisation is not glamorous, but it was hard not to notice that SMPTE made its standards library free to download this year, and that the shared concepts of the cinema package and the Interoperable Master Format already exist. A provider that takes in material from many content owners and delivers to many platforms sees that fragmentation every day, and is well placed to argue for, and help implement, common access-service profiles.
The lesson of the evening is not that television needs smart glasses. It is that the hard parts of accessibility are mostly operational. They are the access components that drop out between ingest and air; the ad breaks without subtitles; the frame-rate conversion that leaves audio description behind; the platform that does not signal an audio track; the absence of data proving that anyone uses it; and the monitoring that stops at the playout server rather than the sofa.
Those are exactly the joins that a managed playout and media services partner touches every day, and where having one accountable team, rather than five suppliers each assuming another is checking, makes the difference. From Leeds, the opportunities we see are these:
Make access-service validation a standard part of ingest and quality control.
Treat subtitle and audio description continuity across breaks, including server-side adverts on free ad-supported channels, as a channel quality measure.
Carry access assets through media processing, including the conversion from 24 to 25 frames per second, as one package.
Extend round the clock monitoring to confirm access services at the platform, and not only at the output.
Support companion-device synchronisation as a by-product of playout.
Combine prepared and live captioning for live channels.
Package personalised versions for streaming and free ad-supported delivery.
Report access-service usage alongside every other audience measure.
None of this requires a technological leap. All of it requires someone in the chain to decide that accessibility is part of the business case for television delivery.
The cinema industry, as it described itself at SMPTE UK, is not short of technology. It has glasses, receivers, applications and standards, and in the Netherlands it has the content too. What it is still working on is making those things reliable enough, and visible enough, that a disabled customer can expect them rather than hope for them. Television is in a surprisingly similar position, with more regulation, better data and far more delivery platforms to go wrong on.
Weronika Kolodziejʼs question for the panel was how accessibility can become an expectation. For those of us who run the chain between the master and the viewer, the answer may be fairly simple: stop treating access services as an extra track to carry, and start treating them as part of what “on airˮ means.
If you would like a second pair of eyes on how subtitles and audio description travel through your own chain, from ingest to the sofa, we would be glad to talk it through. It is a conversation between people who run channels, not a sales call.
As ever, if there is anything we can help with, you know where we are.