The problem to solve
When two or more people speak, the viewer needs to know who said what. In most scenes the picture helps: you can see whose lips are moving. But shots change, people speak off screen, voices overlap, and some viewers cannot hear the voices at all. Subtitles therefore need a way to mark speaker changes that works without much effort from the reader.
There is no single standard. The usual methods are described here, along with where each tends to work well. Which one a project uses is typically set by its style guide.
Dashes
The most common method in subtitles, particularly for dialogue between two people in one cue, is a dash at the start of each speaker’s line:
- Are you coming?
- In a minute.Advantages: it is compact, widely understood, and needs no extra words. Limits: it shows that the speaker changed but not who is who, so it relies on the viewer being able to work that out from the picture or the context.
Conventions vary. Some guides use a dash only for the second speaker, some for both, and some add a space after the dash while others do not. Whichever is used, it should be applied consistently. A dash can also mean an interrupted sentence, which is a different job; see the notes in the article on subtitle punctuation.
Name labels
When the speaker may be unclear, a name in brackets or capital letters can be placed before the line:
[MARIA] I did not know that.Labels are common in captions for viewers who are deaf or hard of hearing, and are most useful when:
- the speaker is off screen,
- several people are in shot and it is hard to tell who is talking,
- the speaker has not yet been named or introduced, or
- a voice comes through a device, such as a radio.
Drawbacks are that they use up space and can interrupt reading. They can also give away information the story is saving, for example a character who has not yet been revealed. In that case a descriptive label (“Man’s voice”) or a neutral one may be preferable. The same bracket style is often used for sounds, discussed in non-speech information.
Color coding
Some systems assign each speaker a color, so lines from the same person always look the same. It can work well where there are a few regular speakers, and it is quick to read once learned. But it has real limits:
- Color alone is not accessible to viewers with color vision differences, so it should not be the only signal.
- Viewers have to learn the code, which is hard in a one-off video.
- It can fail if the viewer uses a device that overrides color or shows only one color.
- Not every format or player supports colored text consistently.
- Readability must be maintained, since not every color has enough contrast against the picture.
Some contexts, particularly some broadcast traditions, use a limited set of colors for this purpose. How color influences viewers more generally is a separate topic, covered in subtitle color. For most ordinary projects, color works best as a supplement to dashes or labels rather than as the only method.
Placement
Another approach is to put the text near the speaker: left of screen for a person standing on the left, right for one on the right. This links the words to the face and can reduce the need for labels. It is common in live captioning for events, where the screen layout can be flexible, and also in some professional subtitling traditions. Drawbacks: it can cover important parts of the picture, it takes extra effort to set up, and it may not be supported by all players or file types. The idea is also touched on in the article on caption displays.
Combining methods
In practice, projects often use more than one method. A typical set of rules might be:
- Use a dash when two speakers alternate in one cue and the picture makes clear who is who.
- Add a name label when the speaker is off screen or unclear.
- Use color or positioning only if the platform supports it and the style guide requires it.
Speaker changes across cues
Where each speaker has their own cue, a dash is often unnecessary because the change in cue itself signals the change. The viewer can see the break. Guides differ on whether a dash is still needed. What matters is that the viewer does not mistake a continuing speech for a new voice.
In a fast exchange, many short lines may be crammed into one cue. That is where reading speed and line length can conflict with clarity. Sometimes it is better to drop a minor interjection than to make a cue impossible to follow.
Overlapping speech
When two people talk at once, most guides prefer to show the more important speech or to put the two lines together with dashes, in the order they are heard. Showing everything may exceed what viewers can read. This is a case where condensing and judgment matter more than any rule.
Automatic labels
Automatic transcription tools sometimes tag speakers, a process called speaker diarization. These labels are typically generic (such as “Speaker 1”) and may be wrong, particularly when voices are similar, people interrupt, or audio quality is poor. A human reviewer should check them before they are shown to viewers.
Choosing a method
Think about the audience and what they need. For a general audience with clear pictures, dashes are normally enough. For viewers who rely on the text for dialogue, labels add clarity. For live events, placement can help. Whatever is chosen, explain the convention if it is unusual, and use it the same way throughout the project.