Editions, Not Translations
Publishers who have run international editions know something that companies entering localization for the first time usually do not: a foreign-language edition is an editorial product, not a translated one.
The distinction is substantive. A translated edition takes the source publication and renders it in another language. An edition takes the source publication's standards, voice, and authority and produces something for a different readership — different story selection, different assumed context, different sense of what needs explaining and what does not.
Video is where publishers now face this question again, usually with a smaller team and less institutional memory than when they built their text editions.
The temptation is to treat video localization as a distribution problem: run the existing video output through translation, publish it in six languages, count the incremental audience. That approach produces content that is technically accessible and editorially foreign — full of unexplained domestic references, assumed political knowledge, and story priorities that make no sense to the target reader.
The publishers doing this well make a smaller number of deliberate choices: which video content genuinely travels, which needs editorial adaptation, and which should not cross borders at all.
Sorting the Video Output
Publisher video output is heterogeneous, and it does not localize uniformly.
Travels well with translation alone. Explainers on international topics, science and health coverage, documentary and long-form investigation with global relevance, business and markets analysis, technology coverage, and anything already written for a reader without local context.
Travels with editorial adaptation. Domestic politics and policy coverage, where the underlying story matters internationally but the assumed context does not. These need added framing rather than a different translation.
Travels poorly. Local news, service journalism tied to a specific jurisdiction, coverage assuming familiarity with domestic institutions and personalities, opinion aimed squarely at a domestic debate, and lifestyle content built around a local market.
Should not travel. Content whose publication is lawful in the source jurisdiction and may not be elsewhere, coverage subject to reporting restrictions that vary by country, and material where identifying an individual carries different legal exposure in different markets.
That last category is not a hypothetical concern for publishers. Defamation standards, privacy law, contempt rules, and restrictions on reporting legal proceedings vary substantially across jurisdictions. A publisher pushing localized video into new markets is publishing there, with the legal exposure that implies. This needs legal input before launch, not after a complaint.
Editorial Standards Applied to Machine Output
Publishers hold themselves to accuracy standards that most organisations do not, and this shapes what an acceptable workflow looks like.
The core problem is familiar but sharper here: machine translation produces fluent, confident output, including where it is wrong. A newsroom reviewer skimming for awkwardness will pass a sentence that reads beautifully and misstates a fact.
The specific risks in journalism:
Attribution collapse. "According to officials," "the company said," "sources familiar with the matter," "allegedly" — the entire apparatus of sourcing and attribution is hedging language, and hedging is exactly what fluent translation smooths away. A report that becomes an assertion has changed from journalism into a claim the publisher cannot support.
Quote integrity. Direct quotations carry a promise of accuracy. A translated quote is by definition not verbatim, and publishers need a settled convention — whether translated quotes are marked as translations, whether the original-language quote is retained, and how paraphrase is distinguished from quotation.
Named entity errors. Names, titles, institutions, and places rendered incorrectly are the most visible errors a publisher can make, and they are systematically preventable with locked terminology.
Numeric drift. Figures, dates, percentages, and currency conventions differ by locale and a wrong number spoken fluently is undetectable by ear. Extract every numeral from source and target and compare them as lists.
Legal terminology. Terms describing legal status — charged, arrested, convicted, accused, under investigation — have precise meanings and different equivalents across legal systems. Getting these wrong creates direct defamation exposure.
The workable standard: machine translation as a first draft, reviewed against source by a journalist who works in that language, with attribution and quote handling checked explicitly rather than incidentally. This is faster and cheaper than translating from scratch, and it is not the same as publishing unreviewed output.
Transparency With Readers
Publishers trade on trust, which makes disclosure a strategic question rather than a compliance one.
The positions that hold up:
Disclose that translation was machine-assisted, in a consistent, findable form. Readers are broadly accepting of this when told and react badly to discovering it. Several major publishers have adopted standing disclosure language and it has not damaged them.
Disclose synthetic voice. If the narration is generated rather than performed, say so. This matters more than translation disclosure because audiences infer a human presence from a voice.
Never synthesise a journalist's voice without their explicit consent, and be careful about synthesising it at all. A reporter's voice speaking words they did not say, in a language they do not speak, is a category of content that needs deliberate policy.
Never synthesise the voice of a source or interview subject. This should be an absolute line. A source consented to being recorded, not to a synthetic version of their voice speaking translated words. Subtitle interview subjects; dub the narration around them.
Provide a correction path per language. Corrections are core to editorial credibility, and a correction applied to the source version but not the six localized versions is a failure of the correction process itself.
That last point is an operational requirement that publishers routinely underestimate. If a fact is corrected, every language version must be corrected, and there must be a system that makes that happen rather than relying on someone remembering.
Syndication and Rights
Publisher video frequently contains material the publisher does not wholly own, and localization can exceed the licence.
Check before localizing:
- Wire service and agency footage, which is licensed with specific territorial and usage terms that translated redistribution may exceed.
- Contributor and freelance agreements, which often grant limited rights that do not contemplate derivative translated works.
- Music licensing, which is territorial and does not extend automatically.
- Interview subject consent, particularly for sensitive stories where the subject agreed to a specific audience.
- Archive and library material with defined usage terms.
- Existing syndication deals that may grant a partner exclusivity in a market the publisher now wants to enter directly.
The last one is a genuine strategic trap. A publisher that syndicates its video to a partner in a market may have granted rights that prevent it from publishing its own localized version there. This needs checking against the actual contracts before a market launch is planned around it.
The Subscription Question
For publishers with subscription businesses, localized video sits inside a larger strategic question: is the goal reach or revenue?
Reach. Localized video as top-of-funnel, freely available, building brand and audience in markets where a subscription product does not yet exist. Measured on audience growth and brand awareness. Legitimate, and it needs to be understood as a brand investment rather than a revenue line.
Revenue. Localized video as part of a subscription proposition in markets where the publisher sells subscriptions. This requires enough localized content to constitute a product — a handful of translated videos is not a reason to subscribe — and it requires payment, customer service, and marketing infrastructure in that market.
Licensing. Localized video sold to partners in markets the publisher will not enter directly. Lower margin, no infrastructure requirement, and it forecloses direct entry later.
The failure mode is doing the first while measuring against the second, then concluding localization does not work. Publishers should decide which model applies per market before launch and instrument accordingly.
A related practical point: catalogue depth determines whether a subscription proposition is viable. Launching a market with twenty localized videos and expecting subscription conversion is optimistic. Building meaningful depth first — including back catalogue, which localizes efficiently in batch and carries no deadline — is what makes the proposition real.
Newsroom Workflow
Fitting localization into a newsroom means fitting it into a production cycle that does not pause.
What works:
Two tiers by speed. A fast tier producing subtitles for time-sensitive content within the news cycle, and a considered tier producing fully localized versions of explainer, feature, and investigative content where the shelf life justifies it.
Terminology maintained continuously, not per project. Newsrooms cover recurring entities — politicians, institutions, companies, places — and a glossary that updates as the news cycle moves is far more valuable than one built once.
A named editor per language, accountable for what publishes in that language. Distributed responsibility produces unreviewed output.
Localization decided at commissioning. Knowing at commission that a piece will run in six languages changes how it is scripted — less assumed context, fewer domestic references, cleaner narration. This costs nothing and improves output more than any downstream intervention.
Back catalogue processed in batch during quiet periods, which uses capacity efficiently and builds the depth that subscription propositions need.
Building the Team
Localized video sits awkwardly in most publishing organisations, and where it reports largely determines whether it works.
Placing it purely in production yields technically competent output that nobody has editorially owned. Placing it purely in the international or commercial function yields content optimised for reach rather than standards. Placing it in the newsroom without dedicated capacity means it is the first thing dropped when the news cycle intensifies.
The structure that holds up has three elements.
A named editor per language with real authority. Not a translator and not a coordinator — someone with the standing to hold a piece back. Diffuse accountability produces unreviewed output, and in journalism unreviewed output is precisely the failure mode that matters.
A small central production capability owning the pipeline, terminology, and delivery specifications across all languages. Duplicating this per language is expensive and produces inconsistency between editions.
Representation at commissioning. Someone speaking for the localized editions in the commissioning conversation, so pieces destined for six languages are scripted accordingly — less assumed context, fewer domestic references, cleaner narration. This is the cheapest quality intervention available and it costs only a seat in a meeting.
On sourcing the per-language editors: journalists who work in that language and understand the publication's standards are far more valuable than translators who do not. The skill that matters is editorial judgement about attribution, sourcing, and framing, which is harder to acquire than language fluency and impossible to substitute for.
Publishers that staff this properly find the localized editions develop their own editorial character over time, which is exactly what distinguishes an edition from a translation.
Frequently Asked Questions
Should translated quotes be marked as translations?
Adopt a consistent convention and publish it. A translated quotation is not verbatim, which sits uneasily with the promise quotation marks make to readers. Common approaches include marking translated quotes explicitly, retaining the original-language quote alongside, or converting to indirect speech. What matters most is consistency and transparency rather than which convention you pick.
Can we synthesise a reporter's or a source's voice?
For reporters, only with explicit informed consent, and it warrants a deliberate policy given that a journalist's voice carries their professional identity. For sources and interview subjects, treat it as a line not to cross — they consented to being recorded, not to a synthetic version of their voice speaking translated words. Subtitle subjects and dub only the narration.
What is the biggest accuracy risk in localized journalism?
Attribution collapse. Reporting language — "according to," "allegedly," "sources said" — is hedging, and fluent translation systematically smooths hedging away. A report that becomes an assertion has changed into a claim the publisher cannot support, and it reads perfectly well, which is why monolingual review does not catch it.
Do we need legal review before publishing localized video in a new market?
Yes. Publishing localized content into a market means publishing there, and defamation standards, privacy law, contempt rules, and reporting restrictions vary substantially. Content lawful in the source jurisdiction is not automatically lawful elsewhere. This is a pre-launch requirement, not a response to complaints.
How much localized content is needed to support subscriptions in a new market?
More than most publishers expect. A handful of translated videos is not a reason to subscribe. Building catalogue depth first — including back catalogue, which batches efficiently and carries no deadline pressure — is what makes a subscription proposition credible. Launching thin and concluding that localization does not convert is a common and avoidable mistake.
Related reading: News Video Translation | Global Content Distribution Guide | Long-Form Video Translation
