Subtitles: The Complete Guide to Creating, Translating and Pricing Them

Subtitles illustrated: a video player with a subtitle bar, surrounded by floating cards of subtitle lines and a translation arrow

Subtitles are the cheapest way to make a video work in a language it was not recorded in. They are also the most misunderstood: teams treat "add subtitles" as one task, when it is really four — transcribe, time, translate, and decide how good the result has to be.

This guide walks the whole chain. Where the detail gets deep, it hands off to a dedicated article; each one is linked in place and listed at the end.

Key Takeaways

Subtitles are four jobs, not one: transcription, timing, translation, and review. Most quality problems come from skipping the last two.
SRT covers almost every use. VTT matters for the web, and burned-in subtitles matter for social feeds that autoplay without sound.
Automatic generation is good enough for a first pass in clean audio. Accents, jargon and several speakers are where it degrades.
Machine translation of an existing subtitle file is fast and cheap; the cost is timing drift and brand vocabulary, both of which a human pass fixes.
Subtitles keep the original audio. When the audio itself is the barrier, dubbing is the next step up — not a better subtitle.

What subtitles actually are

A subtitle file is a plain-text list of timed lines. Nothing more. Each entry carries a start time, an end time, and the text to show in between. That simplicity is why subtitles travel so well between tools — and why so much can go wrong in the handover.

Two words get used interchangeably and should not be. Subtitles assume the viewer can hear the audio and just needs the words. Captions assume they cannot, so they also mark speakers and relevant sound. If accessibility is the reason you are doing this, you want captions, and the extra information is not optional.

For the formats, quality levels and terminology in full, start with what subtitle translation is.

Step 1 — Getting the text

Every subtitle starts as a transcript. You can type it or let speech recognition do the first pass, which is what almost everyone does now. The realistic expectation: clean, single-speaker audio transcribes very well, and everything else needs correcting. Background noise, strong accents, overlapping speakers and product names are the four things that reliably break automatic transcription.

The workflow, tool by tool, is in how to auto-generate subtitles.

Step 2 — Timing

Timing is where amateur subtitles announce themselves. A line that appears half a second late, or hangs on screen after the speaker has moved on, reads as sloppy even when the words are perfect. Two rules carry most of the quality: no line longer than roughly two lines of text, and no line on screen for less than about one second.

Automatic tools produce timings from the audio, so the first pass is usually close. It stops being close after translation — see the next step.

Step 3 — Translation

Translating a subtitle file is not translating a document. The text has to fit the same time slot in the new language, and languages do not keep their length. German runs long against English, so a line that fitted comfortably now overruns its slot and has to be cut. That is the single most common defect in translated subtitles, and no engine solves it for you.

There are three routes: machine translation of the file, a full platform that transcribes and translates in one pass, or a human subtitler. Which one fits depends on how public the content is. The comparison, with tools, is in the step-by-step guide to translating subtitles; the AI-specific detail is in AI subtitle translation.

Step 4 — Deciding what it is worth

Subtitle work scales with minutes, and the price range between "good enough for an internal update" and "goes on the website" is wide. Rather than guess, work backwards from where the video will be seen. Internal training content and raw social clips survive a machine pass. Anything a customer sees before they buy does not.

The numbers, including what full video platforms cost against subtitle-only tools, are in what subtitle translation costs.

Where subtitles stop being the answer

Subtitles leave the original audio in place. That is their strength — the speaker stays the speaker — and their limit. A viewer reading a translation is doing two things at once, and for anything longer than a couple of minutes, attention drops.

If the video is a product demo, a training module or a sales pitch aimed at a market that does not share your language, the honest comparison is not between good and bad subtitles. It is between subtitles and dubbing. We wrote the trade-off up separately in AI dubbing vs subtitles, and if you want to see how the tools in that category actually differ, the alternatives comparison puts eight of them on the same seven facts.

How Dubly handles subtitles

Worth saying plainly, because it explains what this cluster is doing on a dubbing company's website: we do not sell a subtitle tool. Dubly translates finished videos into 40+ target languages with the speaker's own cloned voice and matching lip movement, and every project exports an SRT file and burned-in subtitles as part of that — two styles, with your own accent colour, on every plan.

So subtitles come out of the same upload as the dubbed audio. If subtitles are all you need, cheaper tools exist and this guide names them. If you are subtitling because dubbing looked out of reach, the pricing article is the one to read.

Where to go next

Start with the step that is actually blocking you. Getting text out of the video: auto-generate subtitles. Turning an existing file into another language: how to translate subtitles. Deciding what to spend: what it costs.

You can also test the full path on your own footage — a short video is free, subtitles included.

Subtitles assume the viewer hears the audio and only needs the words. Captions assume they cannot hear it, so they also identify speakers and describe relevant sound. Accessibility requirements refer to captions, not subtitles.

SRT for almost everything — every editor and platform reads it. VTT if the video plays in a browser player that needs styling or positioning. Burned-in subtitles when the video autoplays muted in a social feed, because a separate file will not be loaded there.

For internal content, usually yes. For anything public, no: automatic transcription misses product names, struggles with accents and overlapping speakers, and the timing needs a pass after translation. Budget the review time rather than the correction time after someone complains.

They make a video watchable in another language, which is not the same as effective. For short clips they work well. For product, training and sales video, dubbing keeps the viewer's attention on the picture instead of the bottom third of the screen.

Yes. Every project exports an SRT file, and burned-in subtitles in two styles with a custom accent colour. Both are included in every plan rather than sold as an add-on.

About the author

Maximilian Engler

Co-Founder | Product