Subtitle glossary: 46 terms for captions, timing and delivery

Subtitle terminology: how the terms fit together

Subtitle vocabulary mixes file-format names, broadcast conventions, accessibility practice and informal shorthand. The same word can mean different things in different rooms: a producer may say caption when they mean a same-language track, while a translator may say subtitle for a translated track. A useful reference does not pretend those usages are identical. It gives each term a clear working definition and notes where context changes the meaning.

The terms also arrive in layers. At the bottom are the container and file formats: SRT, WebVTT, TTML and ASS/SSA. Above them sit timing concepts such as cue, timecode, frame rate and reading speed. Accessibility adds SDH, closed captions, open captions and audio description. Production adds transcription, diarization, sidecar delivery, burn-in and archiving. Learning the layers separately makes a technical conversation much easier to follow.

Use the definitions as a shared vocabulary, not as a substitute for the specification you are delivering against. When a client says a file must be 'broadcast compliant' or 'SDH ready', ask which document and which language guide they mean. The words are a starting point for that question.

Subtitle formats explained: SRT, WebVTT, TTML and ASS

SRT is the common plain-text format: numbered cues, timestamps with a comma before milliseconds, and text. WebVTT is the W3C format for web video, with a WEBVTT header, timestamps that use a dot, and optional cue settings for alignment and position. VTT is usually shorthand for WebVTT or its .vtt extension, though people sometimes use it loosely for any web caption file. TTML is an XML family used in broadcast and streaming pipelines, and ASS/SSA is a script-based format with styling and effects that remains common in anime and fansub workflows.

A sidecar file travels beside the video and can be switched on, styled or replaced. A burn-in is rendered into the picture and cannot be turned off. Hardsub and softsub describe those two states, although the words are informal and sometimes refer to how a release was assembled rather than to a formal specification. Muxing combines video, audio and subtitle streams into one container; it does not automatically make the subtitles selectable.

Character encoding is a separate layer from the subtitle format. UTF-8 can represent the scripts most productions need, while a BOM at the start of a file can confuse a parser that expects the first characters to be WEBVTT. If a file opens in a text editor but shows nothing in a player, check the header, the timestamp separator and the encoding before rewriting the cues.

Subtitle timing terms: timecode, frame rate, CPS and gaps

A cue is one timed event: an in-point, an out-point and the text between them. Timecode labels a position in the programme, often as hours, minutes, seconds and frames. Frame rate determines how long one frame lasts, so a timecode in frames and a subtitle file in milliseconds are two views of the same timeline. Drop-frame timecode skips selected frame numbers to stay aligned with real time at rates such as 29.97 fps; non-drop-frame timecode counts every frame and gradually diverges from the clock. The fractional rate 23.976 fps is 24000/1001, a common film-origin rate.

Reading speed is usually expressed as characters per second, or CPS: the number of characters in a cue divided by its duration in seconds. Characters per line, or CPL, measures width instead of pace. Minimum duration and maximum duration bound how briefly or how long a cue may appear. A gap is the space between one cue and the next; an overlap is the opposite, where two cues share time. Overlaps can be deliberate for simultaneous speakers, but they are often a sign that a timing edit went wrong.

A line break is a manual newline inside a cue, not the automatic wrapping a player performs. Manual breaks are useful when they follow the grammar, and harmful when they split a name or a unit. Cue timing and text layout are related but distinct: a cue can fit the character limit and still be unreadable because the break lands in the wrong place.

Accessibility caption terms: SDH, WCAG and audio description

A subtitle usually carries dialogue, often as a translation. A caption usually represents the audio more fully, including speaker labels and meaningful sounds; the distinction is a convention rather than a strict technical rule, and usage varies by country and platform. A closed caption can be switched on or off, while an open caption is always visible, either burned in or delivered as a track the viewer cannot disable. SDH means subtitles for Deaf and hard-of-hearing viewers and typically includes non-speech information that ordinary subtitles omit.

A forced subtitle appears for foreign dialogue or essential on-screen text even when the main subtitle track is off. Some pipelines call this a forced narrative. Audio description is a separate narrated track that describes important visual action for viewers who are blind or have low vision. WCAG is the W3C's Web Content Accessibility Guidelines; its success criteria include captions for prerecorded and live media, and audio description for video.

These terms describe different jobs, and a single file rarely does all of them. A translated subtitle track is not automatically an SDH track, and an SDH track is not automatically audio description. Asking which audience the deliverable serves is more useful than arguing about the label.

Captioning workflow terms: sidecar, burn-in, muxing and archive

Automatic speech recognition, or ASR, turns audio into text and often supplies timestamps. Diarization goes further by partitioning the audio by speaker, which helps a reviewer assign speaker labels. A transcript is the text of what was said, usually without cue timing, though it may carry timestamps or speaker names. A speaker label identifies who is talking when the picture does not make it obvious.

Word-level timing attaches start and end times to individual words, enabling karaoke highlighting and precise alignment. Caption display styles have their own names: pop-on cues appear as complete blocks, paint-on cues reveal words progressively, roll-up cues scroll upward as new lines arrive, and karaoke captions highlight words in time with the audio. The first two are common in pre-recorded work; the live styles favour different production tools.

Delivery terms describe where the text lives. A sidecar file is separate from the video; a burn-in is part of the image; muxing puts streams into one container; encoding converts text or media into a coded representation. Archiving is not a format term, but it belongs in the vocabulary because a project without a master file and a transcript is difficult to revise.

Quick reference: 46 terms defined

The table below is the compact reference. Read a definition in the context of the section above it, because several terms have a narrow technical meaning and a broader everyday one. Where a platform or style guide defines a term differently, the delivery document wins for that project.

How to read a definition without over-generalising

A definition describes a concept; it does not supply the number your platform expects. For example, CPS is a measurement, not a universal ceiling, and minimum duration is a setting that changes by guide and by language. The same is true of line length: CPL is a count, while the limit depends on the destination and the script. Use the term to ask a precise question, then consult the relevant specification or test in the player.

Be equally careful with format names. SRT and WebVTT are often described as interchangeable because both carry timed text, but their syntax differs in small ways that matter to strict parsers. A file renamed from .srt to .vtt is not automatically a WebVTT file. Likewise, a softsub release is not simply a video with subtitles; it implies a selectable track or sidecar that a player can handle.

Using the vocabulary in a real review

When you open a file, name what you are looking at before you change it. Is this a cue with a malformed timestamp, an overlap between two cues, or a line break in the wrong place? Is the problem the character encoding, the header, or the player's rendering? A shared vocabulary turns a vague complaint into a checkable defect.

A practical review uses only a few terms at a time: cue, timecode, CPS, CPL, gap, overlap and speaker label cover most structural problems. Check the timing first, then the text, then the layout. The subtitle validator can flag structural issues, the SRT editor lets you inspect and repair cues, and the guides to subtitle reading speed and subtitle line length explain the limits you are applying. Start from the VoiceToTextFree home page if you need a first transcript.