Create SDH Subtitles for Deaf and Hard-of-Hearing Viewers

What SDH adds to a subtitle track

Ordinary subtitles assume the viewer can hear. They carry the words and leave everything else to the soundtrack. SDH, short for subtitles for the deaf and hard of hearing, fills that gap with three kinds of extra text: who is talking when the picture doesn't make it obvious, descriptions of sounds that affect the story, and notes about music or lyrics.

The difference shows up in small moments. A doorbell that makes a character turn round, a voice from the next room, a song whose lyrics comment on the scene, a sudden silence: a hearing viewer gets all of it for free, while someone relying on text misses it unless it is written down. SDH is close to what many people call closed captions; the practical difference is mostly how the text is delivered, which the section on files below covers. Captions versus subtitles explains how the two terms are used in different countries.

Before you start

• A video in MP4, MOV, M4V, WebM or MKV. Convert anything else first.
• A dialogue source. AI subtitles produce one from the speech; an SRT or VTT from a translator or an editing app works just as well.
• A few house rules. Decide now whether cues go in square brackets or parentheses, whether speaker names are in capitals, and how you mark music. Consistency matters more than which convention you pick.
• Time for one complete watch with the sound up. SDH is written by listening, not by reading a transcript.

The AI draft needs a Google sign-in and uses credits: a 25-minute episode costs 25 credits (2.5¢). Typing, editing, exporting and rendering are free.

How to create SDH subtitles

1. Open the editor, choose your video and run AI subtitles with the subtitle language left on the spoken language, or switch AI off and use Import SRT / VTT.
2. Watch from the beginning with the sound on. Whenever a speaker can't be identified from the picture, click their line and type the name before the words, for example MAYA: Where were you?
3. When a sound matters, pause on it, press Enter to add a line at the playhead and type the cue, such as [door slams] or [phone buzzing].
4. Where music starts and carries meaning, add a line describing it, such as [tense piano music]; for sung lyrics, type the words with a music note at each end.
5. Drag the edges of each new line on the timeline so the cue stays up long enough to read and ends when the sound stops mattering.
6. Clear the warnings in the line list: overlaps, lines that flash past and lines too fast to read.
7. Export SRT or VTT for a player that accepts caption files, or open Style, pick a legible preset and render an MP4 with the SDH burned in.

Conventions worth following

No single SDH rulebook is shared by every broadcaster, streaming service and festival, but the common ground is wide. If you have no client style guide, these are sensible defaults:

Speaker IDs

Name in capitals followed by a colon, used only when the speaker is off screen, hard to see, or two people alternate quickly. Labelling every line is clutter.

Sound cues

Lower case in square brackets, short and concrete: [glass shatters] rather than [a loud noise can be heard]. Describe the sound, not your reading of it.

Music

Mood or source when it matters to the scene: [upbeat pop music], [radio playing jazz]. Skip a background bed that carries no meaning.

Lyrics

Typed out when audible and relevant, with a music note at the start and end. If the note symbol doesn't appear in the preview with your font, put [singing] before the lyric instead.

Manner of speech

A short bracket before the words when tone changes the meaning: [whispering], [sarcastic], [in Spanish].

Silence

When a sound stops abruptly and that matters, say so: [music stops].

For a longer catalogue of cue wording and the judgement calls behind it, read how to caption sound effects and music.

Making room for the extra words

SDH adds text without adding time, and that is where most drafts go wrong. The AI subtitles here are deliberately short, around 32 characters on a single line and no longer than about 2.6 seconds each, so a name prefix usually fits. A sound cue, though, normally needs a line of its own.

If a cue lands in the middle of dialogue, give it a separate line just before or after the speech instead of squeezing it into the sentence. Split breaks a long line at the playhead so each half gets its own timing; Merge joins two fragments so a name prefix appears once. The editor flags any line shown for under 0.8 seconds and anything faster than 21 characters per second, a common ceiling for comfortable reading. A line longer than 42 characters wraps onto a second row by itself, and more than three rows is flagged, which is a strong hint to split. Splitting and merging subtitles walks through both tools.

Where the editor stops

Two SDH habits depend on formatting individual lines, and this editor styles the whole track at once. Italics for an off-screen voice or a voice on the phone can't be switched on for one line, because italic is a setting for every subtitle; use a bracketed tag such as [on phone] or a speaker ID instead. Some guidelines also move a caption to the top when on-screen text fills the bottom. Here position (top, middle or bottom, plus an offset) is chosen once per video, so pick the area that is clear most of the time; moving subtitles to the top helps when the bottom is busy throughout.

The AI writes none of the SDH layer for you. It transcribes speech and nothing else: no names, no sound descriptions, no music notes. Speech recognition can also produce words during music or long silences that nobody actually said, so delete anything you can't hear; Whisper hallucinations explains why it happens.

As a file or burned into the picture

SDH reaches viewers in two forms, and one session in the editor gives you both.

• As a file. Export SRT or VTT and upload it to YouTube, Vimeo, a streaming platform or your own web player. Viewers switch it on and off, which is what closed means. If a distributor asks for SCC, TTML or another broadcast format, convert the SRT with a dedicated tool; the editor exports only SRT and VTT.
• In the picture. Render an MP4 and the SDH becomes open captions that appear in every player, including feeds and messaging apps that ignore caption files. High Contrast, yellow text on solid black, is a strong starting preset for legibility.

If you're unsure which one a platform wants, the closed caption generator covers the switchable route and open captions the burned-in one. For a plain dialogue file without the SDH layer, the AI subtitle generator is the shorter path.

SDH in another language

Translated SDH is a two-pass job. Ask AI subtitles for the dialogue in one of the 21 translation languages, check it, then write the sound cues and speaker IDs directly in that language. Translation works from the video's speech, so cues already typed in English can't be carried across automatically; add them after translating, not before. Translated lines are timed by text length rather than by each word, so expect to nudge more boundaries than you would for same-language subtitles.

About the subtitle editor on this page

Opens

MP4, MOV, M4V, WebM, MKV videos your browser can decode

Subtitles from

AI (from the speech, optionally translated into one of 21 languages), an SRT or VTT file, or typing them in

You download

An MP4 with the subtitles drawn into the picture at the original resolution and frame rate, plus SRT and VTT files

Cost

Editing, styling, rendering and SRT/VTT export are free, with no account and no watermark. AI subtitles cost 1 credit per minute (5-credit minimum per video); signing in with Google gives 100 free credits and accounts get 10 free credits a day

Privacy

The video never leaves your device and is rendered in your browser. Only for AI subtitles is the audio sent, over HTTPS, and it is deleted within 30 minutes of delivery

Rendering needs

Chrome or Edge 94+, Safari 16.4+, or Firefox 130+ on a computer. Editing and SRT/VTT export work in any current browser