Subtitling Rates: What Changes the Price of Video Subtitling, and AI Subtitles vs Professional Subtitling
Subtitling rates are usually quoted per minute of video, and the price depends on how much is said, whether a transcript or script already exists, how many languages you need, whether you need SDH or burned-in subtitles, the deadline and how technical the content is. Automatic AI subtitles are fine for internal drafts; published video needs human review. DEMA Solutions provides both, with your terminology.
Who this guide is for
This guide is for the people who own the video and the budget: marketing managers launching a product film in six markets, training and L&D managers rolling out e-learning modules to every subsidiary, and internal communication managers who publish town halls, webinars and CEO messages for staff across Europe. You may never have bought subtitling before, or you may have bought it once and been surprised by the invoice, or by the number of corrections your colleagues in Germany or Spain sent back.
Your questions are practical. Why does one quote come in per minute and another per word? Can we simply switch on the automatic subtitles of our video platform and translate them? And how do we make sure the product names in the subtitles match the manual and the website? The sections below answer each in turn, without quoting prices, because a price without the factors behind it tells you very little.
What subtitling actually involves
It helps to know what you are paying for. Professional subtitling is not one task but a short chain of them, and each link has a cost.
The first step is transcription: capturing what is said in the source language, accurately and completely. If you already have a final script that matches the edit, this step shrinks to a check. If the video is an unscripted interview, a panel discussion or a workshop recording, transcription is real work. The second step is translation or adaptation into each target language. Subtitles are constrained writing: the translator has to keep the meaning and tone while fitting the text into a limited number of characters and seconds, which often means condensing rather than translating word for word. The third step is segmentation and synchronisation, also called spotting or timing: deciding where each subtitle begins and ends, how a sentence is split across two lines and two subtitles, and making sure each subtitle stays on screen long enough to be read and changes at the right moment relative to shot changes.
Around that core sit review, where language, timing and readability are checked in context with the picture, and delivery, either as a separate subtitle file or encoded into the video. This is exactly how our subtitling services are structured: transcription, translation and synchronisation, followed by review and delivery or encoding.
A subtitle can be linguistically correct and still hard to watch. Lines that are too long, subtitles that flash past faster than anyone can read them, or text that covers an on-screen label all damage the viewing experience even when every word is right. That is why the quality of subtitle localization depends as much on timing and reading speed as on the translation itself.
What changes the price: the main cost drivers
When two quotes for the same video differ widely, the reason is almost always that the providers assumed different things about the factors below. Knowing them lets you compare like with like, and often lets you reduce the cost without reducing quality.
Video length and speech density. Runtime is the starting point, but density matters as much. A three-minute product animation with a voice-over every few seconds carries far less text than a three-minute interview in which two people talk quickly and interrupt each other. Dense, fast speech takes longer to transcribe, needs more condensing in translation and produces more subtitles to time. Long silent passages, music sequences and on-screen demonstrations without speech reduce the effort.
Transcription versus an existing script. If you can supply a final script or an accurate transcript that matches the edit, the provider can skip or shorten transcription. Be honest about how final it is: a script written before filming, followed by an ad-libbed shoot, is a reference rather than a transcript, and checking it line by line can take almost as long as transcribing from scratch.
Number of languages. Transcription and timing are done once for the source language. Translation is done once per target language. A common approach is to create a master subtitle file in the source language, with timing approved, and use it as a template for every translation. This keeps the cost of multilingual subtitles closer to a fixed set-up plus a translation cost per language, rather than repeating the whole process for each market.
SDH and closed captions. Subtitles for d/Deaf and hard-of-hearing viewers carry more than dialogue: speaker identification, relevant sound effects and music cues. Deciding which sounds matter to understanding, and describing them consistently, is additional work, and conventions differ by language and platform. If accessibility is part of the requirement, say so at the start; our page on SDH subtitles explains what the deliverable includes.
Burned-in subtitles or a separate file. A sidecar file (SRT or WebVTT are the most common for web and e-learning) is delivered once and uploaded to your player or platform, which can then show the language the viewer selects. Burned-in, or open, subtitles are rendered permanently into the picture, which is useful for social media autoplay or for players that cannot load subtitle files, but it requires a separate encoded video for every language and a new export whenever a subtitle changes.
Turnaround. A normal schedule lets the same linguist translate and a second person review. A rush deadline may mean splitting a long video between several translators and adding a consistency check, which costs more and carries more risk.
Technical terminology. A product demo for a packaging line, a training video on a medical device or a compliance briefing contains terms that must be exactly right and identical to those in your documentation. Research and terminology checks take time unless you supply a glossary.
Revisions and late edits. Changes to the wording after approval are routine. Changes to the video after subtitling are not: if the edit changes, timings shift and every language has to be re-synchronised. Locking the picture before subtitling starts is the single most effective way to avoid paying twice.
Finally, there is the choice between subtitles and voice-over or dubbing. Spoken translation of a corporate video usually costs considerably more than subtitles, because it adds scripting for lip or time sync, voice talent and studio work. For most training, product and corporate video, subtitles are the more economical route, and many companies reserve voice-over for a small number of flagship films.
How subtitling rates and quotes are structured
Quotes for video subtitling are most often expressed per minute of runtime, because runtime is easy to measure and roughly reflects the effort. A good quote, however, is more than one number multiplied by minutes, and reading it carefully tells you what the provider has assumed.
Expect to see separate lines for transcription in the source language, translation per target language, and synchronisation or template creation, plus optional lines for SDH, review by a second linguist, encoding of burned-in versions and rush delivery. When you supply a script or an existing subtitle file, translation may instead be charged per word, the same way a document is priced, because the text already exists and its length is known. Many providers also apply a minimum charge per language or per file, which matters if you send many very short clips: ten one-minute videos may cost more than one ten-minute video because each file has to be prepared, checked and delivered individually. Before you compare those lines, it helps to know whether you actually need SDH, captions or plain subtitles for each video and platform.
When comparing quotes, check four things. Is transcription included, or is the provider assuming you supply a transcript? Is there a review step by a second person, and does it happen in context with the video or only on a text file? What file formats will you receive, and are burned-in versions included for each language? And how are later changes charged, both wording changes and edits to the video? Two quotes that answer these questions differently are not comparable, however close their per-minute figures look.
AI auto-subtitles: where they help and where they fail
Automatic speech recognition has improved enormously, and every major video platform and conferencing tool now offers automatic subtitles at no extra cost. Machine translation can then turn them into other languages in seconds. For some uses this is entirely adequate. For others it creates risks that are invisible until a customer, an employee or an auditor sees them.
Where AI subtitles help. They are useful for internal recordings that are watched once, such as meeting recordings and informal team updates, for making a long video searchable, and as a first draft that a person then corrects. On clear, scripted speech by a single speaker with good audio, the raw transcript can be quite accurate, and starting from it rather than from a blank page saves real time.
Names and terminology. Speech recognition writes what it thinks it hears. Product names, model numbers, brand names, acronyms and people’s names are exactly the words it is most likely to get wrong, because they are rare in its training data. It will happily turn a product range into a common word, spell a colleague’s surname three different ways, or replace a precise technical term with something that sounds similar. Machine translation then translates the error, so a mistake in the source transcript becomes a mistake in every language.
Timing and reading speed. Automatic subtitles are usually segmented by pauses in the audio, not by meaning or by how long a viewer needs to read. The result is subtitles that split a sentence in an awkward place, stay on screen for a fraction of a second, or carry three lines of text while the picture changes twice. Translated subtitles make this worse, because many languages need more characters than English to say the same thing, and the timing was never adjusted for them.
Accents, noise and overlapping speech. Non-native speakers, regional accents, factory noise, music beds and two people speaking at once all reduce recognition quality sharply. These are common conditions in exactly the videos companies produce: trade-fair interviews, plant tours and recorded webinars.
Sensitive content. Some videos carry statements that must be exactly right: safety instructions in a training video, product claims in a marketing film, figures in an investor message, HR and compliance content for employees. A recognition error that changes “do not” into “do”, or a translation that turns a cautious claim into a promise, is a liability. Confidentiality matters as well: uploading an unreleased product film to a free online subtitling service may expose it to terms of use you have not read.
Accessibility. Automatic subtitles do not include speaker identification or relevant sounds, so they are not SDH. If your video must be accessible to d/Deaf and hard-of-hearing viewers, raw automatic captions are a starting point at best. Our article on the European Accessibility Act requirements explains which organisations the law covers.
A hybrid workflow: AI first draft, human review
For many companies, the sensible answer is neither fully automatic nor fully manual, but a hybrid workflow that uses machine output where it saves time and puts a qualified person wherever errors would be visible or costly.
It starts with the source language. An automatic transcript is corrected by a linguist against the video, using your glossary, so that names and terms are right before anything is translated. The same person checks segmentation and timing, producing a clean master subtitle file whose timing is approved once. That file becomes the template for every language. Each translation can then be produced by a linguist working directly, or by machine translation controlled by your glossary followed by full post-editing, depending on how visible and how sensitive the video is. Finally, every language is reviewed in context with the picture, because a subtitle that reads well in a spreadsheet may still be too long for the shot it sits on.
This approach is the subtitling equivalent of machine translation with post-editing for documents: the machine produces a draft, a qualified linguist makes it publishable, and the level of human effort follows the risk. Our article on when post-editing is good enough discusses where that line falls for written content, and the same reasoning applies on screen. If your own team already generates automatic subtitles, a professional subtitle review of language, segmentation, synchronisation and readability can also be bought as a service on its own.
What to send for an accurate quote
The more complete your request, the more accurate the quote and the fewer questions come back. Before asking for prices, prepare the following:
- The video files, or links to them, in their final edit, with the total runtime and the number of separate files.
- Any script, transcript or existing subtitle file in the source language, with a note on how closely it matches the final edit.
- Target languages and markets, including language variants, and which languages are needed first.
- The deliverable: subtitles, SDH or both; sidecar files (and which format your player or learning platform accepts) or burned-in versions; any style rules from the platform.
- Your glossary and reference material: product names, approved terms, previous translations, and links to the manuals or web pages the video refers to.
- Deadline, publication date and review process: who in your company or in each market approves the subtitles, and how much time they have.
The same logic applies to documents, and our guide on what to send a translation supplier covers it for technical manuals in more detail. For video, one extra point matters more than any other: confirm that the edit is final. If it is not, say when it will be, and ask the provider to quote the cost of re-synchronising after late changes.
Keeping terminology consistent across video, manuals and website
A product video rarely stands alone. The same product has an operating manual, a datasheet, a web page and perhaps a training course, and customers notice when the subtitles call a component one thing and the manual calls it another. For a technical product, inconsistent terms are more than a style problem: a service technician who sees one term in the training video and another in the maintenance manual has to guess whether they mean the same part.
The answer is to treat subtitles as part of the same multilingual content as everything else. A shared glossary with your approved product names and technical terms in every language should be applied to subtitles just as it is to documentation; our guide to keeping terminology consistent across translators and suppliers explains how to build and maintain one. Subtitle text can also be stored in the same translation memory as your other content, so that a safety sentence translated for the manual is reused, unchanged, when it appears in a training video. That memory should belong to you, which is the reasoning behind owning your translation memory.
The website is usually where the video is published, which makes it the third leg of the same consistency problem. If you are planning both, our article on website translation cost explains how the same glossary and memory keep the site and its video content aligned and make every update cheaper.
How DEMA Solutions handles subtitling
Our subtitling work follows a three-step core of transcription, translation and synchronisation, with review in context of the video and delivery either as separate subtitle files or encoded into the media. Projects can include subtitle translation, closed captions, SDH for d/Deaf and hard-of-hearing viewers, subtitle review and audio or video transcription. We subtitle corporate presentations, training videos, e-learning modules, product demonstrations, marketing video, interviews and webinars.
For SDH, conventions for speaker labels, sound descriptions and music are agreed at the start of the project, and the track is reviewed for accessibility and readability; where appropriate, we work with associations supporting d/Deaf and hard-of-hearing viewers to test it. Translation follows an ISO 17100 certified process with a second qualified linguist revising every translation, as described on our quality page. Your glossary and translation memory belong to you and are delivered in portable formats, so the terminology in your subtitles stays aligned with your manuals and website. Our teams in Italy, Serbia and Spain coordinate linguists in every market you publish in.
Where TUBE fits
For companies that publish video alongside documentation and web content, TUBE keeps the language assets behind all of it under your control. Your glossary and translation memory control machine translation, every approved translation goes into a memory you own and can export, and leverage reporting shows how much of each update comes from text you have already approved.
Frequently asked questions
Most providers quote per minute of video runtime, often with separate lines for transcription, translation per language, synchronisation, SDH, encoding and rush delivery. When a script or subtitle file already exists, translation may be charged per word instead. Minimum charges per file or per language are common.
Closed captions and SDH usually cost more than standard translated subtitles for the same video, because they add speaker identification and relevant sound information and need specific review. The difference depends on how much non-dialogue audio matters to understanding.
For internal recordings and first drafts they can be. For published marketing, training, safety or compliance video, automatic subtitles should be corrected and reviewed by a person, because names, technical terms, timing and reading speed are where they most often fail.
Usually yes. A source-language subtitle file with approved timing can serve as a template for every language, so transcription and synchronisation are not repeated. Check that the file matches the final edit of the video.
A separate file lets viewers choose the language and is easy to correct. Burned-in subtitles are always visible, which suits social media and players without subtitle support, but each language needs its own video export and any correction means re-encoding.
More languages, dense or fast speech, poor audio, technical terminology without a glossary, SDH requirements, short deadlines and changes to the video after subtitling has started.
Yes, if the same glossary and translation memory are used for both. That keeps product names and technical terms identical across video, documentation and website.
Plan your multilingual video project
If you are budgeting subtitles for product, training or corporate video, send us the runtime, the languages you need and a sample clip. We will tell you which factors drive the effort in your case, where an existing script or subtitle template reduces it, and which videos need full human subtitling rather than a reviewed automatic draft.
How can we help?
Tell us what you need and we’ll get back to you with the solutions.