Subtitles vs dubbing: how to sell your course in several languages

You recorded a course in your language and now you want to sell it in others. The practical question is simple: do you subtitle it or dub it? There is no single answer, but there are clear criteria for choosing. This article lays them out: what the evidence says about language and buying, what each technique offers, what it might cost (with a hypothetical calculation, labelled as such) and why human review and accessibility matter more than they seem to.

Why language matters

In July 2020 CSA Research published a study of 8,709 consumers in 29 countries. According to its press release, 76% of online shoppers prefer products with information in their own language, 40% will never buy from websites in other languages, and 65% prefer content in their language even when its quality is poor. This is general consumer data, not training data, so read it as a signal rather than a promise: a course is a more considered purchase than a catalogue item. Still, it points in one direction: people who understand the offer in their own language buy more easily.

The same study adds something useful: people tolerate imperfect translation more than is usually assumed. That does not license publishing without review (more on that below), but it does mean you can start with a reasonable version and improve it with use.

Subtitles: cheap, fast and accessible

Subtitling means translating the transcript and showing it in sync with the original video. Its advantages are concrete:

• Cost and speed: text is translated and checked far faster than audio. Fixing a typo means editing a line, not regenerating a clip.
• Fidelity: the learner hears the real instructor, with their own intonation and pauses.
• Accessibility: captions are required by WCAG 2.2, success criterion 1.2.2 (level A) for prerecorded audio in synchronized media. That criterion also expects captions to identify who is speaking and to convey meaningful sounds, not only dialogue.

On learning, the evidence is nuanced. A 2023 study in Smart Learning Environments with 79 secondary school students in Bangkok compared English videos with no subtitles, automatic subtitles and edited Thai subtitles. Comprehension was higher with automatic subtitles than with none (p = 0.01), the difference from edited subtitles was not significant, and cognitive load did not differ. The sample is small, from a single school and without a pretest, so it cannot show that subtitles always help. What it does suggest is that, for a video in a language the learner only half masters, written support does not get in the way.

The drawback is attention: the learner has to read while watching. In a lesson with a practical demonstration (a movement, a surgical gesture, a recipe), looking at the text means looking away from the instructor.

Dubbing: three levels, not one

When people talk about AI dubbing they mix three different things:

1. Generic synthetic voice: the script is translated and read by a catalogue voice. It is the cheapest option and the least personal.
2. Voice cloning: the translated audio keeps the speaker's identity and tone. ElevenLabs documents that its dubbing aims to retain the speaker's identity and emotional tone, detects multiple speakers (up to 32 per file) and supports 90+ languages. It requires consent from the person whose voice is cloned.
3. Lip-sync: besides changing the voice, the mouth in the video is adjusted to match the new audio. It is the most immersive and the most delicate, because any mismatch is noticeable.

One warning from ElevenLabs' own documentation deserves to be taken seriously: the dubbed video it produces is a high-quality preview, not a production master, and for final use it advises replacing the audio track of your original video with the dubbed audio. It also offers human-verified dubs as a separate service. In other words, even the vendor separates an automatic draft from a professional deliverable.

What it might cost: a hypothetical calculation

Prices change, so the method matters more than the figure. As of 14 October 2026, the ElevenLabs API pricing page lists dubbing v1 at $0.33 per minute with a watermark and $0.50 without, and dubbing v2 at $2.20 per minute, excluding taxes.

Suppose, purely as an example, a 10-hour course (600 minutes) that you want to offer in one more language. At $0.50 per minute that is about $300; at $2.20, about $1,320. Multiply by every additional language. Then add what the vendor figure leaves out: your time or a reviewer's to listen and correct, translation of written materials, and re-dubbing every time you update a lesson. This is an illustrative calculation, not a quote, and real cost depends on the vendor, the plan and the volume.

Subtitles are usually a fraction of that, and the gap widens with course length. Dubbing, by contrast, is paid once per language and per version of each lesson.

When each one pays off

A practical rule of thumb, without dogma:

• Subtitles first if your content is technical and visual, if your audience already watches video in other languages with subtitles, if you have not yet validated demand in that market, or if budget is tight.
• Dubbing with a cloned voice if you mostly sell to an audience that is not used to reading while watching, if the content is mostly spoken explanation with little on screen, or if the instructor's voice is part of the course's value.
• Dubbing with lip-sync if the instructor appears on camera and closeness is part of the offer (mentoring, personal development, clinical training with the instructor in frame), and if you can review the result carefully.
• Both when it makes sense: dubbed audio with same-language subtitles, and original-language subtitles for anyone who prefers the real voice.

A reasonable middle path is to start with subtitles in the languages with the most potential demand and dub only the lessons that convert best or are watched most.

Human review is not optional

Machine dubbing or translation can fail in ways the learner does not detect but that damage the course: a technical term mistranslated, a figure read backwards, a joke that falls flat, a wrong dose. In medical or legal training, the error is not merely cosmetic.

An example of real limits: in YouTube's automatic dubbing, generated dubs cannot be edited (to fix a bad translation you change the original video's language so the dubs regenerate), a video may be ineligible if it exceeds 120 minutes, has little speech or is spoken too fast, and managing dubs is only possible on a computer. The platform lets you review before publishing if you choose manual publishing, and you should.

A sensible review flow: a native speaker who knows the subject listens to the full audio of the critical lessons; someone checks terminology against a glossary; and every version carries a visible status (translated, dubbed, reviewed, published) so nothing reaches the learner without passing through it. In training, a published error is a sold error.

Publishing several audio tracks

If YouTube is your delivery channel, multi-language audio lets you upload your own audio tracks to a single video. According to its help page, it is available to creators with access to Advanced features, files must be audio-only and roughly the same length as the video, and the track defaults to the viewer's preferred language, which they can change in the player. If an automatic dub already exists for that language, you must delete it before uploading yours.

Keep in mind that a track you upload is a dub you produced and reviewed yourself: the feature does not generate it.

Accessibility: dubbing does not replace subtitles

It is a common mistake to think that once there is a dub, subtitles are redundant. Captions serve deaf and hard-of-hearing people, and WCAG requires them for prerecorded audio (criterion 1.2.2, level A). For people who cannot see the video, criterion 1.2.3 asks for an audio description or a media alternative, also level A; it does not apply, however, when the audio already conveys all the important information in the video.

So dubbing removes the language barrier, not the disability barrier. Each language version needs its own captions in that language, not just those of the original. If your course is already well captioned at the source, that work is your starting point: the transcript is the raw material for translation.

How it fits into LearnSuite

In LearnSuite every lesson has a source version and derived versions per language, with statuses (translated, dubbed with lip-sync, reviewed, published) and a human reviewer who approves before publishing. Dubbing with the instructor's face and voice uses a cloned voice with signed consent, and runs on a per-academy budget.

Conclusion

Subtitles and dubbing do not compete: they solve different problems. Subtitles are the cheap, accessible, faithful base; dubbing, with a cloned voice and lip-sync, is an experience upgrade worth having where demand is proven and you can review the output. Start with the first, measure, and dub what really justifies it.