What text expansion is
When a piece of text is translated, the result is rarely the same length as the original. Text expansion is the term for a translation coming out longer than its source. The opposite, contraction, also happens, and which one you meet depends on the languages involved. In many fields, such as software and packaging, expansion is a design problem because boxes are a fixed size. In subtitles it is sharper still, because space is limited and so is time: a viewer has only a few seconds to read.
It is difficult to give reliable percentages, because the amount varies with the pair of languages, the type of speech and the translator’s style. Anyone quoting a precise figure for a language pair should be treated with caution. What can be said with confidence is that expansion is common when translating from English into many European languages, and that some other pairs tend to go the other way.
Why translations run longer
Several different causes contribute.
Grammar that adds words. One language may need an article, a preposition or a pronoun where another gets by without. A language that leaves out the subject of a sentence may become longer when translated into one that cannot.
Compound and inflected forms. Some languages combine ideas into long single words. Others use endings to express what English does with separate words. Either way, the written length changes, even if the meaning stays the same.
Missing equivalents. If the target language has no short term for a concept, the translator may need a phrase. A single term in the source can become a short description.
Politeness and register. Some languages require longer polite constructions. Forms of address can add words or syllables, as discussed in forms of address.
Writing systems. A script that packs more meaning into each character, as in Chinese, Japanese or Korean, can use far fewer characters for the same content, though characters are displayed wider. A translation out of such a language into an alphabetic one often expands in character count.
Spoken-language habits. Speech contains repetition, hesitation and broken sentences. A faithful translation may tidy these into complete sentences, which are longer than the broken speech.
Why it is a problem in subtitles
A subtitle line has a maximum width, usually expressed in characters or screen percentage, and most style guides keep the display to two lines. The relevant limits are discussed in line length and the two-line limit. Viewers also need time to read, and the rate they can comfortably read at is covered in reading speed. The time a subtitle can stay on screen is set largely by the speech it accompanies and by the shot changes, as explained in display duration.
If a translation is longer than the source, there are three pressure points at once. There is more text to fit in the same space, more to read in the same time, and no extra time to be had when the speaker has already finished. The reading speed rises, and if it passes what viewers can handle, they miss part of the picture or give up on the text.
How writers cope
A skilled subtitler does not simply cut at random. The usual strategy is to preserve meaning while reducing length.
Condense. Remove words that add little: repeated phrases, fillers such as “well” or “you know,” direct address that is obvious from the picture, and explanations the viewer can see. The craft is covered in text condensation.
Choose shorter words. Where there are alternatives, pick the shorter one that fits the tone. A longer formal word may be replaced by a plain one, as long as the character voice does not change.
Restructure the sentence. Reordering clauses can sometimes cut words and also provide a better place for a break.
Spread across cues. If the speech is spread over a long stretch, the translation can use two sequential subtitles instead of one crowded one. This depends on the subtitle timing, and cues should not appear so fast that they flicker or leave no cue gaps between them.
Use the picture. If the screen shows what is meant, such as a gesture or a visible object, the line may not need to spell it out.
Drop what is least important. If something must go, it should be the least important element for the scene, such as a secondary remark. The viewer should never lose the main point.
Adjust the timing. Where space around a cue is free, the cue can start slightly early or end slightly late within the limits of good practice, giving viewers more time. This must not cause the subtitle to appear before the speaker or to overrun a following shot.
What not to do
Shrinking the font or adding a third line to make text fit is rarely a good solution. Small text hurts legibility, and a third line covers the picture. Speeding up the reading rate beyond what the style guide allows trades one problem for a worse one. Another trap is to leave out meaning without noticing, especially in subjects such as health information, where precision matters and omissions can mislead. Where accuracy is critical, a qualified reviewer should check the result.
Planning for expansion
Teams can reduce the problem by planning:
- Set line and speed limits per language, not one limit for all languages.
- Leave room in the schedule for condensation, since a short translation is usually faster to produce than a well-condensed one.
- Where the source text can be controlled, such as a scripted program, keep sentences simple and avoid dense phrasing that will multiply when translated.
- Test samples early, in the real player.
- Keep a record of agreed shorter terms in a term glossary so the choices stay consistent.
Machines and length
Automated translation tends to be literal and does not know the subtitle limits. A machine draft frequently produces lines that are too long, and human editors must then condense it. Treat any automatic output as a draft for editing. Style rules differ among broadcasters and platforms, so use the guidelines supplied for your project.