Subtitles, Captions and CEE Localization for Video

Since 28 June 2025 the European Accessibility Act has applied to the services in its scope, and the technical yardstick it points at is EN 301 549, which for video defers to WCAG success criteria including captions on prerecorded media. In practice that ended one common shortcut. A caption burned into the picture with no time-coded sidecar is not an accessible caption, because nothing can read it, resize it or switch it off. What we deliver, and in which formats: Time-coded files in the format the destination actually consumes. SRT and WebVTT for web and platforms. TTML in the IMSC 1.1 profile where a player demands it. EBU-STL for Polish and European broadcast delivery. SCC where a US partner still needs CEA-608. SDH is produced separately from standard subtitles, with speaker identification and sound events, because they are two files for two audiences and merging them serves neither well. The craft that automated output does not have: Reading speed is a constraint, not a preference. We work to seventeen characters per second for adult audiences and thirteen for anything a child will watch, forty-two characters per line, two lines. A subtitle does not cross a shot change unless it lands within a few frames of it. Polish rendered from English source runs materially longer, arriving with more characters and the same number of seconds, so the real work is condensation: deciding what the sentence is for and cutting the rest. That is a writer's decision. It is why we do not run machine translation, pass a human over it and call the result localisation. Languages we take and languages we hand off: Polish, English, Czech, Slovak, Hungarian, Romanian, German. Those we do with translators who work in category and whom we will name to you. Anything outside that set we will not pretend to review. We introduce you to a vendor who can, rather than subcontracting blind and putting our name on work nobody here can read. Category constraints that change the file itself: Pharmaceutical advertising in Poland carries mandatory reminder wording that has to be present and legible in every localised version, including the vertical crop, which is where it most often disappears. Medical device claims stay tied to the approved wording per market instead of being freshly retranslated each time somebody makes a new cut. Automotive consumption and emissions figures differ by market and by homologation, so a localisation that reuses the source figures is a compliance error wearing a translation's clothes. We ask for the per-market claim set before starting, and if it does not exist yet, we wait for it. What we refuse: We do not auto-generate captions and sell you the correction pass. We do not localise into a market where nobody on the client side can approve the language, because an unreviewed language version is a liability you have paid us to manufacture. We do not do dubbing or lip-sync voice replacement; we record voice-over and we know good dubbing houses. We do not do live captioning. Who should not hire us: Anyone who needs forty languages by Friday. Anyone treating captions as a box ticked for a legal file rather than something a deaf viewer will genuinely use, because our questions will feel like an imposition. Anyone with no source script and no appetite to pay for transcription first. How it runs: We take the master, the source script where one exists, the per-market claim set and your glossary. We transcribe and time to picture, translate, then run the compliance check against the claim set per market before quality control on the files themselves. You get a review link where your local reviewer comments on a timed line instead of emailing about the bit near the end.