Why one set of rules never fits every language
Subtitle guidelines are often written with one language in mind, and then applied to others. That works only up to a point. Languages differ in how they build sentences, how long their words are, how they show politeness, and even in which direction they are written on the screen. A subtitle is a small box of text with strict limits on space and time, so every one of those differences shows up quickly.
This page gives a broad overview. Several of the topics have their own articles on this site, and they are mentioned briefly here rather than repeated. The aim is to help you see why subtitlers working in different languages make different choices, and why a subtitle file that looks fine in one language can need extra care in another.
Grammar and word order
Word order is the first difference many people notice. English usually places the verb early in a clause, close to the subject. Some other languages put the verb at the end of certain clauses, or build long chains of modifiers before the noun they describe. German subordinate clauses and Japanese sentences are common examples. When a subtitler translates a spoken line into such a language, the meaning may only become clear at the end of the sentence, and that affects where the text can be broken.
This matters for how to split a sentence across subtitle lines. Good practice is to break at natural phrase boundaries, so that a reader does not have to hold half a thought across a line break. In a language where the key word arrives last, there may be fewer natural places to break, and the translator may need to restructure the sentence instead of translating it word for word.
Other grammar features have similar effects:
- Dropped subjects. Some languages routinely leave out pronouns because the verb ending shows who is acting. A translation into a language that needs the pronoun adds words, and a translation out of it may have to work out the subject from context.
- Grammatical gender. Many languages give every noun a gender. A line that avoids naming a speaker’s gender in one language may be forced to reveal it in another.
- Plurals, cases and tenses. Languages mark these in very different ways, which can make a short phrase in one language need a longer one in another.
Length of text
Languages do not take up the same amount of space to say the same thing. A translation is often longer or shorter than its source, and the difference depends on the pair of languages involved and on the kind of speech. Words in some languages are long compound forms, while others use many small function words. Text expansion is the usual name for translated text running longer than the source, and it is a routine problem when subtitle space is limited.
Because viewers need time to read, a longer translation cannot simply be squeezed into the same time slot. The writer has to condense the meaning, which is the skill described in text condensation. Dropping repeated words, hesitations and filler is common, but the subtitle still has to keep the point of the line. The limits that apply are discussed in line length and reading speed, and those limits are not the same for every language, partly because of the next topic.
Scripts and writing systems
A script is the set of characters a language is written in, and scripts differ in more than appearance.
Alphabets and syllabaries. Latin, Cyrillic and Greek scripts use letters for sounds. Some scripts, such as Japanese kana, use symbols for syllables. Japanese is usually written with a mix of kana and characters borrowed from Chinese, so one line may contain several types of symbol.
Character-based writing. Chinese writing uses characters that each carry meaning, and a single character can stand for what takes a whole word in English. This is why subtitles in Chinese, Japanese and Korean often have far fewer characters per line than subtitles in alphabetic languages. The characters are also usually displayed wider, so the number that fits on a line is limited by width as well as by count.
Direction of writing. Arabic, Hebrew, Persian and Urdu are written from right to left. That affects alignment, punctuation placement and the way numbers and foreign words are mixed into a line. These display problems are covered in right-to-left scripts, which gets its own article here.
Spaces between words. Not every language puts a space between words. Chinese and Japanese do not, and Thai separates phrases rather than individual words. A computer cannot simply break a line at a space, so the system or the writer has to decide where a word ends, a task known as word segmentation. This too has its own article.
Accents and special letters. Many languages use letters with marks above or below them. If a subtitle file is saved or opened with the wrong character encoding, those letters can appear as broken symbols. That is a technical issue rather than a language one, but it falls hardest on languages with more than basic letters.
Politeness, register and tone
Some languages require speakers to choose between formal and informal ways of addressing someone. French, Spanish and German each have this choice built into the word for “you”, and Japanese and Korean carry politeness through verb endings and vocabulary. English has no such built-in marker, so a translator going into English has to show the relationship through other means, such as word choice, titles and tone. Going the other way, the translator has to pick a form for every pair of speakers and keep it consistent unless the story gives a reason to change. The topic is explored in forms of address.
Register is wider than politeness. It includes slang, regional speech and idioms. Each of these is tied to a place, an age group or a period, and a literal translation can come out odd or even nonsensical. The article on idioms and slang looks at the choices, and the one on regional dialects covers how place-specific speech is handled. Wordplay is the hardest case, since a pun depends on the sounds and meanings of two specific words, which rarely line up in another language. See wordplay translation for how translators weigh the options.
Words left in another language
Characters in a film may speak more than one language, sing in one, or drop in a phrase from another. Whether to subtitle these, and how to mark them, depends on the audience and on what the production intended. Sometimes the point of a scene is that the audience does not understand a character. The question of untranslated speech is separate from translation proper, and it has its own article.
What this means in practice
Differences between languages mean that no single rule about line length, reading speed or line breaks will work everywhere. Professionals usually work from style guidelines written for the specific target language, and a human who reads the target language is the best check on whether the result is natural. This is one reason why human reviewers still matter even when machine translation supplies a first draft. Automated tools often handle common language pairs reasonably but can stumble on less common ones, on politeness, and on anything that depends on culture. Cultural considerations affect many of the choices described above.
A few practical points follow:
- Plan extra time for languages with different scripts or directions, since display and encoding checks add work.
- Agree on terms, names and forms of address early, so they stay consistent across a long project.
- Test the finished subtitles on a real player, in the target language, before release.
- Have someone fluent in the target language review the result, especially for health or legal material, where mistakes can matter.
Requirements and conventions for subtitles vary by country, broadcaster and platform, so treat this overview as general background rather than a rulebook.