Accessibility captions under the European Accessibility Act
A selectable subtitle track (SRT/VTT), not burned-in text
Since 28 June 2025 the European Accessibility Act applies to consumer services in the EU — including video. If your videos carry speech, they need captions, in the language of each market you serve, delivered as a track the viewer can turn on. SublyAI generates that track from your video automatically.
What the act actually requires
In Czech law it is Act No. 424/2023, transposing EU Directive 2019/882 and measured against EN 301 549 (WCAG 2.1 AA). The parts that concern video:
- Captions for every pre-recorded video that contains speech (WCAG 1.2.2, level A)
- Audio description for pre-recorded video (WCAG 1.2.5, level AA)
- Captions in the language of each market — English captions are not enough for a German storefront
- SDH quality: sound effects and speaker identification, not just dialogue
- A hidden track the viewer can toggle, enlarge and read with assistive technology. Burned-in text fails this: it cannot be turned off or resized.
- An accessibility statement published on the site
What SublyAI does today
- Speech to text, automatically. Deepgram Nova-3 transcribes with word-level timestamps.
- 54 languages for both transcription and translation — the limit of our speech model, not a marketing number.
- When the video is already in the language your audience needs, transcribe in that language — no translation step, same price.
- SRT, VTT and TXT export, free, without burning anything in. This is the selectable track.
- Speaker identification (diarization), free. SRT, VTT and TXT can mark who is speaking; you fill in the names, because no model can know them.
- Your video never leaves your browser. Only the extracted audio goes to the AI, and it is deleted afterwards.
- Already have a caption file? Import your own SRT/VTT and burn it in for free.
- 60 minutes free on registration, no card required.
What we do not do yet
Stated plainly, because this decides whether the track is fully SDH-compliant today:
- SDH sound cues ([music], [door slams]) — planned. No model identifies them reliably enough, so we would rather leave them out than invent them.
- Dubbing / synthesized voice-over — planned
- Audio description — planned. It needs a written description of the visual content, not just captioning.
We also do not replace an accessibility audit. We produce the caption track; the accessibility statement, keyboard navigation, contrast and alt text are yours.
Roadmap
- SDH sound cues — [music], [applause] inside the exported track
- Dubbing — synthesized voice in the target language
- Audio description — narration of what is on screen
No dates. We ship when it works; the changelog is the source of truth.
Who the act applies to
Micro-enterprises (under 10 employees and under €2M turnover) are exempt. If that is you, the act does not bind you directly — but it binds your clients: the e-shop selling into Germany, the school with e-learning, the company publishing product videos. Those are the ones who need captions.
Nothing on this page is legal advice. For your specific obligations, read the act or ask a lawyer.
Slovakia
Our Czech interface is readable to Slovak speakers, and Slovak is one of the 54 languages we transcribe and translate. We do not maintain a separate Slovak UI — the Czech one serves both markets.
Questions
- Does SublyAI make my video comply with the EAA?
- It produces the caption track the act asks for. Compliance also covers audio description, an accessibility statement, keyboard access and contrast. We handle captions only.
- Should I burn the subtitles in or export a track?
- Export SRT or VTT. Burned-in text cannot be toggled or enlarged, so it does not satisfy the requirement.
- Do I have to translate the video too?
- No. If the video is already in the language your audience needs, transcribe in that language and export. Translation costs the same as transcription.
- How much does it cost?
- 60 minutes are free on registration. After that it is pay-per-use and the minutes never expire.
- Do the captions say who is speaking?
- Yes. Speaker identification runs automatically and is free. Exported SRT, VTT and TXT can mark each speaker; you fill in the names, because no model can know them. Sound cues such as [music] are not there yet.
- Is my video uploaded to your servers?
- No. Processing runs in your browser; only the extracted audio is sent to the AI and it is deleted after processing.
Put a caption track on one video
Upload it, pick the language, export SRT or VTT. Free minutes on registration.
