Accessibility work is a production discipline

Captioning, subtitling, timed text, and audio description make screen stories usable by audiences who cannot access every part of the original audio or image. These are not generic text-generation tasks. They combine language, timing, editorial judgment, cultural knowledge, technical delivery, accessibility standards, and careful quality control. AI can accelerate transcription, speaker segmentation, translation drafts, shot analysis, or first-pass description, but the final deliverable still has to communicate the work accurately and appropriately. Job titles include caption editor, subtitle editor, subtitler, timed-text specialist, localization quality-control operator, accessibility editor, audio-description writer, audio-description quality-control specialist, and media accessibility coordinator. Search those established titles alongside AI-assisted, automated, machine translation, speech recognition, or media localization rather than expecting one universal AI captioning job title.

Know the difference between the deliverables

Captions represent spoken dialogue and meaningful non-speech audio for viewers who are deaf or hard of hearing. W3C explains that captions include dialogue, speaker identification, and sounds needed to understand the media. Subtitles commonly translate or transcribe dialogue for viewers who can hear the program, although naming conventions vary by market and platform. Subtitles for the deaf and hard of hearing, often called SDH, include relevant sound information and speaker cues. Audio description adds narration for significant visual information that is not available from the soundtrack alone. W3C identifies actions, characters, scene changes, and on-screen text as examples. Each asset has different users, timing constraints, style rules, and technical formats. A professional must confirm the requested deliverable instead of treating every text track as interchangeable.

Standards turn good intentions into measurable quality

Accessibility teams work against defined requirements. The FCC's television captioning rules describe four quality components: accuracy, synchronicity, completeness, and placement. WCAG 2.2 requires captions for prerecorded audio in synchronized web media under Success Criterion 1.2.2, subject to its stated exception. Platform and broadcaster style guides add rules for reading speed, line breaks, speaker identification, punctuation, sound descriptions, forced narratives, shot changes, positioning, and language-specific conventions. The correct guide depends on the client and destination. Build a habit of creating a delivery checklist before editing. Confirm language, audience, frame rate, source timecode, runtime, file format, line and character limits, reading speed, profanity treatment, speaker labels, music handling, and required quality-control report.

What timed-text work looks like

A timed-text specialist receives picture, audio, script or dialogue list when available, a technical specification, and a deadline. The work may involve transcribing, translating, spotting in and out times, choosing line breaks, identifying speakers, describing meaningful sounds, resolving overlaps, checking shot changes, and exporting one or more files. The editor then watches the program in real time and performs technical and linguistic quality control. Automatic speech recognition can create a useful starting point, but names, accents, whispered dialogue, overlapping speech, fictional terms, music, and noisy scenes remain common failure points. Scripts can also differ from the final edit. Treat every reference as evidence, not unquestionable truth, and reconcile it against the delivered picture and audio.

What audio-description work looks like

Audio-description writers identify visual information that is essential to understanding the story, then write concise narration that fits around dialogue and important sound. The goal is not to describe every object or interpret every emotion. It is to communicate useful visual facts in the time available while respecting tone, genre, character knowledge, and narrative reveals. The script moves through editorial review, timing, pronunciation research, recording, mixing, and quality control. AI vision or language tools may suggest objects, actions, or draft phrasing, but they can miss context, infer unsupported motives, confuse characters, reveal information too early, or overload a quiet dramatic moment. Human reviewers must compare every description with the actual sequence and the approved brief.

Use AI as a draft and detection layer

Useful assistance includes speech-to-text, diarization, terminology extraction, translation memory suggestions, reading-speed flags, line-length checks, overlap detection, shot-boundary detection, and comparison of a timed-text file with an approved script. For audio description, computer vision may help create a shot inventory or flag on-screen text. These tools should operate inside a documented workflow with approved data handling. Preserve the source file, tool version, prompts or settings when relevant, editor changes, and final approver. Ofcom's 2025 access-services report notes that providers are exploring AI-assisted subtitling while also emphasizing careful assessment and ongoing human oversight. That is the practical model: automation can increase capacity, but quality ownership remains with trained people.

Human quality control is not optional

A polished file can still be wrong. Quality control includes watching the complete program, not only scanning text. Check whether meaning was preserved, all necessary dialogue and sounds are present, timing follows the intended audio, text remains readable, captions do not obscure important on-screen information, and technical metadata matches the delivery. For translated subtitles, review idiom, register, character voice, cultural references, continuity, and names. For audio description, verify visual facts, timing gaps, pronunciation, mix balance, and consistency. Keep a correction log so recurring problems can improve future workflows. If the deadline does not permit proper review, raise the risk rather than silently shipping an unverified machine output. Accessibility users should not become the final quality-control department.

Localization requires language and cultural expertise

Fluent language skills are necessary but not sufficient. Screen dialogue is constrained by time, reading speed, character, humor, genre, and what the audience can already see. A literal translation may be accurate word by word but fail as a subtitle. The U.S. Bureau of Labor Statistics describes translators as preserving concepts, style, and tone while using glossaries and terminology databases; those principles apply to audiovisual work even though every timed-text role does not map neatly to one occupational category. Work in your strongest language direction and be honest about proficiency. Maintain project glossaries, approved character names, recurring terminology, pronunciation notes, and client decisions. AI translation should be treated as a suggestion that a qualified language professional edits against the picture and source meaning.

Learn the technical side of delivery

You should understand frame rates, timecode, drop-frame conventions where applicable, spotting, forced narratives, closed versus open captions, sidecar files, embedded tracks, character encoding, and the difference between source and distribution masters. Common text formats include SRT, WebVTT, TTML-based formats, and platform-specific packages, but a job posting may require proprietary tools or delivery systems. Do not claim that one file works everywhere. Learn to validate a file, inspect it as text, compare timestamps, and test playback in the intended environment. Keep version names unambiguous and never overwrite the approved master casually. Technical confidence helps you distinguish a language error from a playback, frame-rate, encoding, or packaging problem.

Build a rights-cleared accessibility portfolio

Use your own footage, public-domain material, or content licensed for the purpose. Create a short caption sample with dialogue, speaker changes, music, and meaningful sound effects. Include the final video, the timed-text file, the governing style assumptions, and a brief quality-control report. Create a second sample in your qualified language pair if you seek localization work. For audio description, choose a scene where visual information matters and produce a script with timecodes; a recorded mix is helpful but not essential for a writing role. Add a case study showing how an automated first pass was corrected. Do not present copyrighted streaming footage publicly without permission, and do not use synthetic voices or performer likenesses unless the rights and portfolio context are clear.

Write the resume and prepare for tests

List languages with honest proficiency and direction, relevant accessibility or localization training, timed-text and media tools, technical formats, genres, and completed work you can discuss. Emphasize accuracy, deadlines, terminology research, quality control, and secure handling. If you used AI, explain the human review rather than claiming perfect automation. Hiring tests may ask you to caption, subtitle, translate, spot, review, or write description for a short clip under a style guide. Read the instructions before starting, record assumptions, and reserve time for full playback. In interviews, expect questions about ambiguous dialogue, reading-speed conflicts, sensitive language, speaker identification, client feedback, data security, and machine-output errors. A careful explanation of tradeoffs is more credible than pretending every case has one universal answer.

Find legitimate work and build a specialty

Search accessibility vendors, localization companies, broadcasters, streamers, post-production facilities, media-technology companies, public institutions, education platforms, and production-company career pages. Verify the legal entity and application domain, especially for remote freelance offers. Do not pay a recruiter for access to a captioning role or complete an excessive unpaid test. Rates and employment structures vary by country, language, format, turnaround, and whether the work includes transcription, translation, timing, quality control, or recording, so evaluate the full scope. Over time, specialize in a language pair, genre, live captioning, audio description, quality control, workflow design, or accessibility operations. The most durable career position combines audience knowledge, language or writing craft, technical delivery, and the ability to evaluate automated output responsibly.

Sources and further reading