A note on dates
Exact dates in subtitle history are often disputed, differ by country, or depend on what counts as the “first” of something. This outline therefore describes stages and general periods and avoids precise years where it could mislead. Different regions followed different paths at different speeds.
Text before sound: intertitles
Early films had no recorded dialogue, so cards of text were shown between scenes to give the story, speech or context. These intertitles were a forerunner of on-screen text, and they could be replaced to suit a different language, which made them an early form of translation for film. They were separate cards, though, not text laid over the picture.
Subtitles on film
Once recorded sound became standard, films needed another way to reach audiences who did not speak the original language. Methods of adding text to the picture were developed over the following decades. Early approaches were physical and labor-intensive: text could be printed or etched onto film prints, which meant each language needed its own copy of the film. Because the text was part of the picture, it could not be turned off, which is the same trade-off now seen with burned-in subtitles.
At the same time, many countries developed traditions of dubbing instead. The split between subtitling countries and dubbing countries is a long-standing pattern, though it is not absolute, and it has been shaped by cost, politics and audience habit.
Captions for deaf and hard of hearing viewers
Subtitles translate speech for people who can hear it. Captions were developed to give deaf and hard of hearing audiences the speech and also the important sounds. Early captioned films and television were typically shown to limited audiences, with limited availability, and the work was done by hand.
For television, a technical breakthrough was the idea of sending caption data hidden in the broadcast signal, so that viewers could switch captions on or off. This is the origin of the term closed captions. Some systems carried the text in a hidden part of the analog picture signal, and in several countries a text-based information service delivered on television sets was also used for subtitles. Viewers often needed a special decoder at first. Over time, many places made decoding a built-in feature of television sets, and rules or policies in various countries pushed broadcasters to caption more of their programs. The details, including when each step happened, vary from country to country.
The contrast between closed captions, which a viewer can switch off, and open captions, which are always visible, dates from this era and remains relevant today.
Digital files and the web
As broadcasting and video went digital, subtitles became data. Instead of being printed on film or encoded in a broadcast line, subtitles could be stored in a separate text file that lists each line with start and end times. A range of subtitle file formats appeared, some simple and some capable of carrying styling, position and language details. Players could load them alongside a video, which made it easy to add or swap languages without touching the picture itself.
This shift changed who could make subtitles. Anyone with a text editor or a dedicated subtitling program could time and export a file. Fan communities created subtitles for material that was not officially available in their language, and professional workflows also moved to digital tools with features for timing, checking reading speed and managing many languages at once.
Online video added further change. People began watching on phones and in public places, often without sound, and platforms added tools for uploading captions. For many viewers, subtitles went from a special accessibility feature or foreign-film convention to something they use as a matter of habit.
Speech recognition and machine translation
Researchers had worked on computer recognition of speech for many years before it became practical for everyday captioning. Early systems handled limited vocabularies or required training for one speaker. Gradually, improvements in methods, computing power and available data made general-purpose recognition much more usable, and it began to be built into video platforms and subtitling tools.
Machine translation followed a parallel path, moving from rule-based approaches to statistical methods and then to neural networks, which are now widely used. Taken together, these developments let a tool produce a timed transcript and a translated version with little manual input.
Automated subtitling is now common in the sense that many services offer it, but it has not replaced people. Automatic drafts still make characteristic mistakes with names, accents, noise and jokes, and many publishers still rely on human reviewers for anything important.
Live captioning
Live programs needed a different approach, because there is no time to prepare text beforehand. Methods developed for live work include fast keyboard-based captioning by trained operators, respeaking, and, more recently, automatic systems. Each brings its own balance of cost, delay and accuracy.
What the history suggests
Three themes recur across these stages. First, each new technology tended to widen access while introducing new limits. Second, standards for readability, such as how much text appears at once, have developed through practice and are still debated. Third, automation shifts where effort is spent more than it removes effort altogether. Where earlier generations typed and timed everything, today’s workers more often correct and refine a machine draft.
Readers who want exact dates for a particular country or technology should consult specialist histories or the relevant national bodies, because this outline deliberately stays general.