Patient Education Video Translation: Review First

Why every translated health video needs a clinical gatekeeper

A patient who misunderstands an instruction about medication timing, wound care or warning signs can be harmed. Machine translation and speech synthesis can produce fluent, confident output that is subtly wrong: a dose expressed in the wrong unit, "before meals" turned into "with meals," or a symptom described with a word patients use for something else. Fluency makes those errors harder to spot, not easier.

That is why the process below puts qualified review at the center rather than at the end. Two kinds of expertise are needed: a qualified medical translator or interpreter who is fluent in the target language, and a clinician who can confirm that the translated version still says what the clinical team intends. Neither replaces the other.

Important: this article is general information about producing translated educational material. It is not medical or legal advice, and it does not replace your organization's clinical governance or language access policies.

Start from a plain-language source

Translation magnifies whatever is in the source. A script full of medical jargon, long sentences and abbreviations produces a translation that is just as hard to follow, and harder to check. Before translating, revise the source to plain language:

Short sentences with one instruction each.
Everyday words, with the medical term introduced once and explained, if patients will hear it from clinicians.
Numbers spoken clearly with units: "two tablets, twice a day, morning and evening."
Explicit actions rather than abstract advice: "call the clinic if your temperature is above the number your nurse gave you."
No idioms, jokes or culturally specific comparisons.

A plain-language script also helps patients who speak English as a first language but have limited health literacy, so the effort pays off twice. Most health systems have plain-language or health literacy guidance; use your organization's standards.

What clinical and language reviewers each check

Split the review so each reviewer knows what they own.

Medical translator or interpreter: Accuracy of meaning, terminology patients actually use, register, grammar, readability for the intended audience
Clinician: Doses, units, timing, warning signs, contraindications, and whether instructions still match current practice
Patient or community reviewer: Whether the video is understandable, respectful and trustworthy to people who will watch it
Accessibility reviewer: Captions, readability on screen, audio clarity

For critical content, an independent back-translation can show the clinician where meaning shifted, since most clinicians cannot read every target language. The article on finding a reviewer for AI translation covers sourcing qualified linguists; for health content, look for medical translation qualifications or experienced medical interpreters.

Culture and language variety matter as much as words

Two patients who "speak Spanish" or "speak Arabic" may not be served equally by the same video. Consider:

Language variety. Written Modern Standard Arabic is widely understood in writing, but many patients speak a regional dialect at home and may find formal spoken Arabic distant. Chinese-speaking patients may read Simplified or Traditional characters and speak Mandarin, Cantonese or another variety.
Health beliefs and practices, such as dietary advice that assumes certain foods, or medication guidance during religious fasting periods.
Family roles in health decisions, and whether a video should address the patient alone or their family too.
The presenter and imagery. Some patients are more comfortable with a presenter of the same gender for intimate topics.

Community health workers and patient advisory groups are well placed to flag these issues. The article on cultural adaptation in video gives a broader framework.

Accessibility for every patient

Translated videos should be as accessible as the originals. Provide captions in the same language as the audio, for deaf and hard-of-hearing patients and for patients watching in a waiting room with the sound off. Keep a printable transcript or handout in the same language, reviewed to the same standard, for patients who want to reread instructions at home. Remember that captions and dubbing do not help blind or low-vision patients understand what is shown on screen; if a demonstration matters, the narration must describe it. If you are weighing a voiced version against captions alone, the subtitles vs dubbing guide sets out the trade-offs. The broader picture is covered in AI and video accessibility.

Language access rules

Many countries have rules about language access in healthcare. In the United States, for example, federal civil rights law, including Title VI of the Civil Rights Act and Section 1557 of the Affordable Care Act, sets language access obligations for many healthcare organizations, and recent Section 1557 rules address machine translation of critical content by requiring review by a qualified human translator. Requirements, interpretations and enforcement change, and they depend on your organization and location. Check current official guidance and ask your compliance or legal team how they apply to you. This is not legal advice.

Example: an insulin pen video for a diabetes clinic

A community diabetes clinic has a seven-minute English video showing how to use an insulin pen, and many of its patients speak Spanish or Vietnamese. The nurse educator first rewrites the script in plain language, with each step as one short sentence and every number spoken with its unit. She produces draft translated transcripts and dubbed drafts in both languages, then sends them to a contracted medical translator for each language. The translators correct terminology and flag one step where the draft said "shake" for a motion the clinic describes as "roll gently." The clinic's diabetes specialist checks doses and timing against the corrected English back-translations. The clinic then shows the final videos to a small group of patients from its advisory panel, asks them to explain the steps back, and adjusts one confusing sentence before publishing.

A review-first production process

Choose a video with stable, current clinical content, approved by the responsible clinician.
Rewrite the source script in plain language and get clinical sign-off on the English version.
Decide languages from your patient population data and interpreter request records.
Produce a draft translated transcript and, if needed, a dubbed draft. Label every file clearly as a draft.
Send drafts to a qualified medical translator, who corrects them and returns a reviewed script.
Have a clinician confirm doses, timing and warning signs, using a back-translation where needed.
Regenerate the video from the reviewed script, or record a human narrator reading it.
Test with patients who speak the language, using a teach-back approach: ask them to explain the instructions in their own words.
Publish with captions and a matching handout, record who reviewed which version, and set a review date.

Limits and risks: why AI output stays a draft

AI translation can be confidently wrong, and patients usually cannot tell. Only qualified review catches this.
Speech recognition errors, such as a misheard drug name or number, carry into the translation.
One synthetic voice may not convey reassurance or urgency the way a trained presenter does. For sensitive topics, a native-speaking human narrator reading the reviewed script may be better.
Languages outside the tool's list, and many regional varieties, are not covered.
Text in the picture, such as labels on medication packaging or on-screen dosage tables, is not translated by dubbing.
Translated videos never replace an interpreter in clinical conversations. They support education; they are not consent, diagnosis or individual advice.

What mydubly can and can't do for clinical teams

mydubly can generate the drafts your translators and clinicians review. You upload a video in the browser and get a transcript, a timestamped transcript and subtitles in the spoken or a target language, or a dubbed version with a translated voice and video. The timestamped transcript lets a reviewer find each line in the video quickly. It supports 21 languages; Arabic output is Modern Standard Arabic and Chinese output is Simplified, which matters for the language variety points above. See the video translator page for the outputs.

It is not a medical translation service. It does not provide human translators or clinical review, does not translate on-screen text, and uses one stock voice for the whole video. mydubly makes no compliance or certification claims. The video file stays on your device, only compressed audio chunks are sent over HTTPS, and results are deleted within 30 minutes of completion; the private video translation page has the details. Patient education videos normally contain no patient information, and you should keep it that way unless your privacy team has approved the workflow. As a cost example, the seven-minute video above costs 7 credits (0.7¢) for a translated transcript in one language, and 350 credits (35¢) for a dubbed draft.

For drug names, doses and clinical terms, the glossary approach in translating names and technical terms gives every reviewer the same reference.

A safe first step

Pick one short, stable patient education video, rewrite its script in plain language and line up a qualified medical translator and a clinician before producing any translation. Run the full review-first process on that single video, including a patient teach-back test, to learn how much review time each video needs. The video translator page explains what the draft outputs look like.

Frequently asked questions

Can we publish AI-translated health videos if a bilingual nurse checks them?

A bilingual clinician is a valuable reviewer for clinical accuracy, but being bilingual is not the same as being a qualified medical translator. For patient-facing materials, pair clinical review with review by a qualified medical translator or interpreter, and follow your organization's language access policy. Some rules specifically require qualified human review of machine-translated critical content. This is not legal or medical advice.

Should patient videos be dubbed or subtitled?

Many patients prefer to hear instructions, and some have limited literacy in any language, so a translated voice usually reaches more people. Captions remain important for deaf and hard-of-hearing patients and for noisy waiting rooms. A good default is a voiced version with captions in the same language and a reviewed printed handout. For sensitive topics, consider recording a native-speaking human narrator reading the reviewed script.

How do we choose which languages to translate first?

Use your own data: the preferred languages recorded for patients, interpreter request logs and community health data for your area. Prioritize languages where many patients rely on interpreters and topics where misunderstanding causes harm, such as medication, post-procedure care and warning signs. Check that a qualified translator is available for each language before starting, because review capacity often limits a program more than production does.

Is it safe to upload patient education videos to an online tool?

Generic education videos, without patients or identifying details, are usually low risk, but follow your organization's privacy and procurement rules. Never upload recordings of real patient encounters unless your privacy team has approved the workflow. mydubly sends only compressed audio chunks over HTTPS and deletes results within 30 minutes of completion, but it makes no healthcare compliance claims, so your team must decide whether it fits your policies.

How often should translated patient videos be reviewed?

Whenever the clinical guidance, medication, device or local process changes, and on the review cycle your organization uses for patient materials. Update the English source first, then re-translate and re-review every language together so no group of patients receives older advice. Display a last-reviewed date on the video page and keep a record of who approved each version.