Three kinds of text, three jobs
A video can have a script, a transcript and subtitles, and all three are made of words. They are easy to confuse because they overlap, but each is built for a different purpose, and using one where another is needed usually goes badly.
In short: a script is written before or during production to guide it, a transcript is a record of what was actually said, and subtitles are short timed pieces of text shown with the picture while it plays.
The script
A script is a plan. It is written by the people making the video, before it is recorded or while it is being shaped. Depending on the type of production, it may include dialogue, narration, stage directions, scene descriptions, camera notes and timing guidance.
Its purpose is to help people make something. It is not meant to be a faithful record of the final product. Lines get changed, cut, improvised or reordered during filming and editing, so the finished video often departs from the script.
Because it exists early, a script is useful to a subtitler for spellings, names and intended meaning. But it cannot be trusted as the final word. Whether to translate from a script or from the audio is its own question, discussed under source material.
The transcript
A transcript is a record. It is produced after the speech has happened, by a person typing it out or by speech recognition followed by correction. It aims to capture what was said, in order.
Transcripts vary in how literal they are. A verbatim transcript keeps hesitations, repeated words and false starts. A clean or edited transcript removes them and tidies grammar. Many also include speaker labels, and sometimes time markers every so often so a reader can find a place in the recording.
Transcripts are meant to be read, searched, quoted or archived. They suit people who cannot or prefer not to listen, and are handy for podcast transcripts and meetings. They have no fixed line length or display timing, since the reader controls the pace.
Subtitles
Subtitles are a presentation. They are tied to the video through timing: each cue appears at a particular moment, stays for a short period and disappears. Because they share the screen with the picture, they must be short and quick to read.
Several limits follow from that:
- Lines are short, and usually no more than two are shown at once.
- Viewers read at a limited pace, so text is often condensed rather than copied word for word.
- Each cue must be timed to the speech and, ideally, to shot changes.
- Line breaks are chosen to help reading, not just to fit space.
Subtitles may also be a translation, or may be captions that include sound effects and speaker identification for viewers who cannot hear. The difference between captions and subtitles is explored in the article on captions versus subtitles.
A side-by-side summary
| Script | Transcript | Subtitles | |
|---|---|---|---|
| Main purpose | Guide production | Record what was said | Present speech on screen |
| Made | Before or during production | After the recording | After the picture is final |
| Faithful to the final video? | Not necessarily | Intended to be | Intended to match, but condensed |
| Timing | Usually none | None or occasional markers | Exact start and end for every cue |
| Length of text | As long as needed | Complete | Short, edited to fit |
| Read by | Cast and crew | Readers who want the full text | Viewers, while watching |
Why they cannot simply replace each other
A script is not a transcript. It may contain lines never spoken and omit lines that were added. Using it as the record of what was said can mislead.
A transcript is not a set of subtitles. A full transcript is usually too long and too fast for viewers to read on screen. It needs splitting into cues, condensing and timing. Condensing dialogue without losing meaning is a skill in itself.
Subtitles are not a transcript. Because they are shortened, they omit words, and each cue is displayed briefly. They are poor for reading as a document or for search, though they can be exported as plain text for rough use.
How they feed each other
In practice the three are linked. A script may supply names and the intended wording. A transcript, whether typed or produced by speech recognition and corrected, becomes the raw material for subtitles. Subtitles can in turn be exported to produce a basic transcript, though this will lack some content.
A common workflow is: obtain the best available text, check it against the audio, divide it into cues through the spotting process, condense where needed, and then time and review.
Choosing what to produce
Ask what the reader or viewer needs. If someone needs to follow a video, they need subtitles or captions. If they want to read, search or reference the content without playing it, a transcript serves better. If a team is making the video, a script is the planning tool. Many projects benefit from more than one, and each should be reviewed on its own terms. Important material, such as health or legal content, should be checked by a qualified human reviewer whichever form it takes.