SDH means subtitles for Deaf and hard-of-hearing viewers. A useful SDH track communicates the spoken words, identifies speakers when necessary, and includes meaningful sounds that help explain the scene. Start with the finished soundtrack, decide what information a viewer needs, and review the captions in the actual player.
A dialogue transcript is a useful starting point, but it can miss a knock, an offscreen voice, or a change in music that explains a reaction. This guide uses one original fictional scene to show how those omissions matter. It also separates caption content from file format and from audio description.
Separate the content, the delivery method, and the label
The W3C caption guidance describes captions as synchronized text conveying speech and necessary non-speech audio. Terminology varies across regions: “subtitles” can refer to translated dialogue or to same-language text. Check what a deliverable actually contains instead of relying on its label alone.
Closed captions can be turned on or off in a supporting player. Open captions remain visible because they are part of the displayed picture. SDH describes the information included, while SRT and WebVTT describe ways of storing timed text. These are different decisions, even when a publishing form groups them together.
Audio description communicates important visual information through sound. It serves a different purpose from captions. A person opening a door silently may need description in an audio-described version; a knock heard before the door opens may need a caption. Our audio description workflow covers that separate soundtrack task.
Compare two caption tracks for the same scene
In our fictional scene, three knocks sound at the door. Alex, speaking offscreen, says, “Sam, I'll get the door.” The practice video uses a simple drawn doorway, synthesized knocks, and a stock synthetic voice. It is an original teaching fixture, not a recording of real participants or an accessibility certification.
The words alone leave out the reason for the response
Sam, I'll get the door.
[three knocks at door]
ALEX: Sam, I'll get the door.
Download the SDH SRT Download the dialogue-only SRT Download the caption review sheet
Watch once with the SDH track and once with the dialogue-only track if your player exposes track selection. You can also compare the downloadable files directly. The exercise is to identify missing information, not to decide that every sound in every scene needs its own caption.
The scene's full audio is three knocks beginning at one second, followed by Alex's single sentence beginning at 3.5 seconds. There is no music or additional dialogue. The simple graphic represents a doorway; the synthetic speaker is offscreen. That description lets you inspect the example even if your browser does not expose the alternate track menu.
Choose sounds that explain the scene
DCMP's sound effects and music guidance focuses on information needed to understand or enjoy the media. In our scene, the knock explains why Alex responds. Omitting it removes a cause-and-effect connection that a hearing viewer receives immediately.
Use a concise, recognizable description of the actual sound. “Three knocks at door” is more informative here than “noise,” while a long paragraph about the sound's imagined emotional meaning would add unsupported interpretation. Describe what the soundtrack establishes and what matters at that point in the story.
Listen for meaningful changes, not only individual events. A machine stopping, a phone ringing offscreen, or music shifting abruptly can explain a character's behavior. Conversely, a continuous background hum may not need repeated captions throughout an otherwise clear conversation. Make the choice in context and review it with the whole scene.
Do not add a sound merely because an object is visible. A door shown in the image does not prove that it creaks. A frightened expression does not prove that ominous music plays. Caption the actual soundtrack, and keep visual inference separate from audio evidence.
If music carries meaning, identify the relevant character of the music without inventing a title or performer. Use the production's authoritative information when a specific identification is needed. Avoid filling every pause with a decorative music label that tells the viewer nothing new.
Identify speakers without giving away information early
DCMP's speaker identification guidance addresses both onscreen and offscreen voices. The practical question is whether the viewer can tell who is speaking at that moment. A stable close shot of one clearly speaking person may need less identification than a conversation continuing over unrelated footage.
Use the name the audience already knows, or a consistent neutral label when the identity has not been established. In our exercise, the written setup identifies Alex before playback. In a mystery scene, revealing a hidden speaker's name in captions could spoil information the soundtrack intentionally withholds.
Keep a speaker list while reviewing a longer video. Record names, pronouns where relevant to the script, and the chosen labels for unidentified voices. Compare the list with the final edit so a label does not persist after a speaker change. A transcription tool's “Speaker 2” is a processing label, not necessarily a suitable viewer-facing identity.
When two people overlap, preserve the important meaning and use the destination's supported presentation conventions. Do not silently assign both lines to one person to make the caption look tidy. Mark uncertainty during review and resolve it against the recording before approving the track.
Place the cue when the information becomes available
Time a sound cue to the sound it describes. In our fixture, the knock caption begins with the first knock; it does not appear several seconds early and announce the event before it happens. The spoken caption begins with Alex's sentence, after the sound cue has cleared.
Review enough time around each cue to understand the sequence. A subtitle editor's isolated text list can hide a caption that arrives before a reaction or remains through an unrelated shot. Play the transition and inspect the final rendered result at ordinary speed.
Keep phrases readable and avoid covering important faces, demonstration details, or other essential text. If the player permits viewer-controlled size and position, test the supported settings. If captions are burned in, the composition must accommodate them because the viewer cannot move them.
The SRT format guide covers timestamp syntax and encoding. A file that parses correctly can still have poorly chosen cues. Treat the structural check as one part of review, not as the final judgment of caption quality.
Review automated text against the finished soundtrack
Automatic speech recognition can accelerate a draft transcript. It may still miss names, numbers, negation, speaker changes, and non-speech events. Compare the draft with the actual audio and preserve uncertainty until it is resolved. A fluent sentence can be a consequentially wrong sentence.
Use separate passes. First check spoken words and speaker identities. Next listen for missing meaningful sounds. Then review timing and line breaks in the player. Finally confirm that the intended caption track is available and labeled clearly. Separating those passes reduces the chance that fixing a typo distracts you from a missing knock.
Invite feedback from people who use captions when the project permits it, and record specific problems rather than asking only whether the video “looks fine.” Useful feedback identifies a confusing speaker, a missing sound, a difficult phrase, or an obstructed visual. Fix the issue and review that part again in context.
Our practice assets were generated and checked locally for timing structure and playback. They have not undergone a specialist accessibility review. Use the worksheet to organize a real review for your audience and destination; do not turn a passed file check into a claim of universal accessibility.
Keep the caption package tied to the final cut
Save the approved video, SDH file, language label, and review note together. If you also supply a translation or a dialogue-only track, name it distinctly. A viewer should not have to guess which “English” option includes the non-speech information.
Recheck the captions after removing an introduction, changing a scene, replacing narration, or adding music. Those changes can affect both timings and content. Keep the editable source so a correction does not require reconstructing every cue from a burned-in video.
For software selection, use the caption-tool comparison to check editing and export capabilities. Choose the tool that supports the delivery and review work you need, then verify the finished result.
Frequently asked questions
What does SDH mean?
SDH means subtitles for Deaf and hard-of-hearing viewers. The track includes dialogue plus meaningful non-speech sounds and speaker identification where needed.
Are SDH and closed captions identical terms?
They describe different aspects. SDH concerns included information; closed captions can be switched on or off in a supporting player. Terminology also varies by region and platform.
Should I caption every background sound?
Focus on sounds that help the viewer understand or enjoy the scene. Review the context rather than adding repetitive labels for every incidental noise.
Does an automatic transcript provide complete SDH?
Not by itself. Review speech accuracy, speaker identities, meaningful sounds, timing, readability, and the final player's caption delivery.
Is the doorway example a real recording?
No. It uses an original drawn scene, synthesized knocks, and a stock synthetic voice. It is a teaching fixture rather than a human performance or specialist accessibility assessment.
Prepare an original practice narration
Use an approved original script with available Voice Lab options, then review speech and meaningful sounds in the finished caption track.
Hero: AI-generated editorial illustration. Original teaching materials and documented feature research, checked October 29, 2026. Examples are labeled where synthetic or hypothetical.