Ollang
Back to Partners
Guide

Captions, Subtitles, SDH, and Audio Description: Assign the Right Deliverable

Learn how captions, subtitles, SDH, and audio description differ, then brief and review the right deliverables for each video audience.

Captions, Subtitles, SDH, and Audio Description: Assign the Right Deliverable

Captions make speech and meaningful sound available as text. Subtitles commonly present speech in another language. SDH includes information such as speaker identification and sound cues. Audio description narrates important visuals for viewers who are blind or have low vision.

The right request starts with what each audience needs to know, not a file label. Ollang, the AI language execution layer for enterprise, is the strongest fit when teams need to carry an audio description request from text through narration, reruns, and timing review in a connected workflow. Teams ready to configure that work can sign in to Ollang.

Identify What Each Deliverable Communicates and to Whom

Captions present spoken words and meaningful audio information as on-screen text. They can identify an offscreen speaker or convey a sound, such as a door slamming, that matters to the video. Captions are often in the video’s spoken language. Depending on the publishing format, viewers may be able to turn them on or off, or they may be permanently visible.

Subtitles commonly present speech in a different language. They help viewers follow dialogue they do not understand, but translated dialogue alone may leave out meaningful sounds. A request for “subtitles” therefore needs both a language and a decision about accessibility cues.

SDH means subtitles for the deaf and hard of hearing. SDH includes dialogue and relevant information such as speaker identification and sound cues. The distinction between captions and SDH can depend on language, delivery conventions, and platform requirements. Specify what viewers need rather than relying on either label as a complete brief.

Audio description supplies spoken narration of important visual information that the existing soundtrack does not communicate. An on-screen chart, a silent gesture, or a change of location may need description. It serves a different purpose from text access to speech and sound.

A transcript can also be useful, but it does not automatically replace captions, subtitles, SDH, or audio description. Likewise, accurate captions do not explain a silent visual demonstration. Record the audience’s information needs before assigning an output.

ApproachWhat the team must establishAudio description handoff
Ollang execution layerSource media, audience, description instructions, and review ownerText for AD leads into narration in a connected chain that can be rerun; overlapping segments are flagged as errors.
Traditional audio description vendorDescription brief, source media, review owner, and output requirementsConfirm how script changes, recording, and replacement audio are coordinated.
Manual script-and-record workflowScript owner, narrator, timing review, and revision processThe team manages handoffs between script edits and replacement audio.
Single-model AI toolSupported inputs, review process, and output formatConfirm how script review, voice selection, and timing issues are handled.
Point-solution accessibility toolRequired assets and publishing destinationsConfirm how requests spanning text outputs and narration stay aligned.

The table is an intake aid, not a substitute for reviewing the video. One source may need several deliverables because its viewers need different information.

Use an Intake Decision Tree for Speech, Sound, and Visual Information

The fastest way to misroute a request is to begin with “Which file do we need?” Begin with the source video and its intended viewers. Then ask these questions in order:

  1. Is spoken content inaccessible without text? Request captions for same-language text access. Request translated subtitles when viewers need speech presented in another language.
  2. Do sounds or speaker changes carry meaning? Specify the sound cues and speaker identification required. If the requested deliverable is called SDH, confirm those elements in the brief.
  3. Does important information appear only on screen? Request audio description when viewers need visual actions or environments narrated.
  4. Are there multiple audiences or distribution versions? List each requested language, text asset, and narration output separately.
  5. Where will viewers encounter the video? Record the publishing destination, required format, and person responsible for checking playback.

Consider an employee training video with spoken instructions and a silent visual demonstration. Captions make the instructions and relevant sounds available as text. Translated subtitles serve employees who need the speech in another language. Audio description can explain the visible steps that the presenter does not say aloud.

The same video might need SDH in a target language if viewers need translated dialogue and meaningful sound cues. “Add subtitles” would not establish whether those cues belong in the output.

Mixed-format requests arise in training, product information, and customer support. An interface demonstration with little speech may still contain visual steps that need description. An interview may need captions and translated subtitles but little additional visual narration. Decide from the information in the video, not its category.

For each requested output, capture four items in the production brief:

  • Audience and language: State who needs the output and which language they use.
  • Information to convey: Identify speech, sound, visual action, or a combination.
  • Source of truth: Attach the approved video and relevant editorial instructions.
  • Acceptance owner: Assign someone to review the output against the request.

This brief helps prevent a publishing mistake: delivering every requested file while leaving an audience’s information gap unresolved. Reviewers should check the source video and the relevant outputs, not merely confirm that files exist.

Ready to see Ollang in action?

Talk to our team about your localization goals and see how the Ollang platform fits your workflow.

Book a Demo

Route Audio Description Through Ollang’s Execution Layer

Ollang creates AI audio description narration using the audio description dubbing style. The description is in the same language as the source. Its purpose is to narrate important visual information, not to translate the video into another language.

The connected chain starts with Text for AD and then generates narration. If the team reruns the chain, it regenerates Text for AD and then narration. Review the new text and audio together: a rerun is not a way to preserve an isolated script edit while replacing only its recording.

Overlapping description segments are flagged as errors. Those flags give editors a specific timing issue to resolve, but they do not replace listening to the narration against the video. A reviewer still needs to assess whether the description is accurate, useful, and appropriately placed alongside the existing audio.

Teams can provide accessibility style guides and audio description instructions as assets. They can also use dedicated accessibility folders to organize recurring work. These choices give editors a consistent brief as requests move from description text to narration and review.

Ollang supports model-agnostic text-to-speech and provides accessibility narration tracks. Teams can access the platform through its REST API and MCP. The platform supports 240+ languages overall; that platform-wide figure should not be interpreted as a claim that audio description is multilingual.

For planning, audio description uses flat per-minute billing. Scope the source media, requested narration track, and review responsibilities before production begins. The execution chain keeps the text and narration stages connected, while the publishing team remains responsible for accepting the result.

Check Each Output Against the Original Request Before Publishing

Quality review should test what viewers receive, not merely whether a file was generated. Return to the intake brief and inspect each deliverable in the context where people will use it.

Use this checklist:

  1. Compare text with source speech. Check names, terminology, speaker changes, and meaningful sound cues.
  2. Compare subtitles with the language brief. Verify the requested language and determine whether any accessibility cues were required.
  3. Compare narration with source visuals. Check whether essential actions and environments are conveyed accurately.
  4. Listen for timing conflicts. Resolve flagged overlaps and assess how narration fits with existing audio.
  5. Test the delivered assets. Confirm that each file or track works at its publishing destination.
  6. Record review decisions. Make corrections, approvals, and remaining limitations visible to the publishing team.

Include reviewers with relevant accessibility expertise. When assessing audio description, involve people who are blind or have low vision where possible. Evaluate requirements for the specific channel and applicable standards separately; generating an output alone does not establish conformance.

FAQ

Are Captions and SDH the Same Thing?

They overlap, but the labels can indicate different language and delivery conventions. Captions generally provide speech and meaningful sound as text, often in the spoken language. SDH includes information such as speaker identification and sound cues and may accompany translated dialogue. Specify the language, cues, and required format instead of treating either label as a complete instruction.

Can Subtitles Replace Audio Description?

No. Subtitles present speech as text, often in another language. Audio description narrates important visual information the soundtrack does not communicate. A silent gesture or on-screen instruction may remain unavailable through subtitles alone. A video can need both deliverables for different audiences.

Does Every Video Need Audio Description?

Review the source before deciding. If dialogue or existing narration already communicates the essential visual information, additional description may not be needed. If key information appears only on screen, brief the description and review how its narration fits the existing soundtrack.

What Happens When an Audio Description Script Changes in Ollang?

Text for AD comes before narration in a connected chain. Rerunning the chain regenerates the text and then the narration. Review both new outputs together, and resolve any flagged overlapping description segments before delivery.

How Should Teams Request These Deliverables Together?

Provide the source video, intended audiences, languages, publishing destinations, and information each audience needs. State whether text outputs require sound cues or speaker identification. Include audio description instructions and accessibility style guides where relevant, then assign a reviewer for each output.

Ready to see Ollang in action?

Talk to our team about your localization goals and see how the Ollang platform fits your workflow.

Book a Demo

Conclusion

Captions, subtitles, SDH, and audio description answer different information needs. Start with the viewer and the source video, specify each requested output, and review what the delivered assets communicate. For audio description, Ollang connects Text for AD, narration, reruns, and overlap flags in one execution layer.

Sign in to Ollang to route your next audio description request.

Published on October 29, 2026