The European Accessibility Act and Your Video Library

If your reaction to the European Accessibility Act is "we should get to that at some point," the point you're picturing has already happened. Directive 2019/882 became applicable on 28 June 2025, and it sets accessibility requirements for audiovisual media services — streaming and on-demand video among them — that most video libraries built before anyone was thinking about this were never scoped against. There's a real second deadline still ahead, in 2030, but it covers a narrower slice of your catalogue than most teams assume, and treating it as a blanket extension is the single most common way this gets mishandled.

The date that already passed

The Act doesn't single out video as its own category — it covers a range of products and services, with audiovisual media services named specifically among them. For a streaming platform, an on-demand video library, or any service delivering video content to EU users, 28 June 2025 is the date the underlying accessibility requirements became applicable, not a future target to plan around. If your back catalogue hasn't been scoped against it yet, the honest framing isn't "we're ahead of a deadline." It's "we're behind one that's already here," which changes the right next move from planning to triage.

The exception that isn't a reprieve

Here's the detail that gets flattened into "we have until 2030" far more often than the text actually supports: services already on the market undergoing no substantial change have until 28 June 2030 to come into conformity. Read that as two conditions stacked together, not one grace period covering everything you currently run.

"Already on the market" means the service existed before the main date — new services, or new video products launched after June 2025, don't inherit this tail at all. They're subject to the requirement now, same as everything else.

"No substantial change" is the condition that quietly disqualifies far more of a working catalogue than teams expect. A service that gets meaningfully updated — and what counts as "substantial" is a judgment call worth getting conservative advice on, not assuming in your own favor — loses the extended runway for that change onward. A back catalogue frozen exactly as it was in June 2025 might genuinely ride the 2030 tail. A back catalogue you're actively re-editing, re-releasing, or substantially updating is a much shakier bet to assume the same protection for.

The practical read: 2030 is real, but it's a narrow lane for content you're leaving alone, not a five-year pass on doing anything about accessibility across your library generally.

Triage before remediation

With one deadline already active and a second one that only protects untouched content, the actual first task isn't fixing videos — it's sorting them. A back catalogue of any real size can't be remediated in one pass, so the sorting question is what buys you a defensible order to work in:

1. What's actively serving traffic right now? The highest-viewed, highest-revenue, or most-linked content is where non-compliance carries the most real exposure, and it's the content most likely to get substantially updated anyway — which pulls it out of any 2030 protection regardless.
2. What's about to be touched for an unrelated reason? Anything scheduled for a re-edit, a re-release, or a refresh is about to lose "no substantial change" status the moment that work happens. Bundling accessibility remediation into a change you were making anyway is close to free compared to a standalone remediation pass later.
3. What's genuinely dormant? Old, low-traffic, untouched video is the actual candidate for the 2030 lane — leave it alone in both senses, don't remediate it yet and don't substantially change it either, since either action changes its status.

That triage turns an undifferentiated "the whole library needs this" problem into three separate queues with three different levels of urgency, which is the only way a back catalogue project like this actually gets started instead of stalling at the size of the number.

Three remediation tiers, cheapest first

Once a video is in the queue to actually fix, the levers available scale in cost roughly in this order, and it's worth working through them in sequence rather than reaching for the most expensive one by default.

Transcription. Getting a plain-text transcript of a video's spoken content is the cheapest, fastest lever, and it's useful even before you've decided how the final captions should look — it gives you searchable, editable text you can check for accuracy on names and terms before anything gets burned onto the video itself.

Captions. Adding auto-timed subtitles is the next tier, and for most video accessibility obligations, it's the one doing the actual compliance work — burned-in captions rendered onto the video itself, timed to the speech, in whichever of 165 supported language codes matches your audio. This is a single automated pass end to end: point the tool at the video, and it transcribes and times the captions in one motion.

Audio description. This is the tier that changes character rather than just cost. Where transcription and captions are each a single automated tool call, audio description — a spoken narration track describing meaningful visual content for viewers who can't see it — isn't a one-click Versely capability. It's a composite: someone writes a description script for what's happening visually in the gaps between existing dialogue, that script gets voiced, and the result gets mixed into the original audio rather than replacing it. It's the most labor-intensive tier precisely because it isn't automated the way the first two are, which is exactly why it belongs last in the sequence rather than first — clear the cheap, high-coverage levers across the whole triaged queue before spending the heavier effort on the videos that specifically need it.

Building the remediation pass in Versely

A concrete path through a queued video, cheapest lever to most involved:

1. Pull a transcript first, even if captions are the actual goal. "Transcribe the audio from this video to text." This runs transcribe_audio and gives you a checkable, editable transcript before anything is rendered permanently onto the video.
2. Caption the video from that same audio. "Add subtitles to this video automatically, source language English." This runs add_veed_captions, transcribing and timing in one pass — for the large majority of a back catalogue, this is the step that actually closes the gap.
3. For the smaller set of videos that need audio description, build it as its own short pass rather than trying to force it through the captioning tool: write the description script for the visual gaps, generate it with generate_speech in a distinct narrator voice from any existing dialogue, then use attach_audio_to_video in mix mode to layer the new narration under the original track without stripping it out.
4. Log which videos got which tier, especially any that were substantially changed in the process — that log is what tells you, later, which parts of the catalogue can still plausibly claim the 2030 lane and which ones already left it the moment you touched them.

The takeaway

The European Accessibility Act's main clock started running in June 2025, not some point still ahead of you, and the 2030 date it's easy to mistake for a full reprieve only protects the content you leave genuinely untouched. Triage the library by what's live, what's about to be edited anyway, and what's truly dormant — then work the cheap, automated levers across as much of the queue as they cover before spending the heavier audio-description effort on the videos that actually need it.