International video metadata often travels farther than the video itself. A broadcaster exports a clip to a partner, an archive sends it to a catalogue, and a newsroom publishes a shortened version in another language. At each stop, people may translate the title and description, but the video’s identity, rights, creators, technical facts and provenance must remain dependable. IPTC Video Metadata Hub offers a common vocabulary for describing these properties across several technical representations; it is not merely a place to store multilingual captions.
If you want to translate IPTC Video Metadata Hub fields, localize multilingual video descriptions, map IPTC VMH 1.7 values into JSON, XMP or EBUCore, or avoid confusing the video’s spoken language with its catalogue language, this guide explains the decisions that matter. We will cover content descriptions, language labels, title variants, usage rights, people, locations, dates, technical parameters, AI-related provenance fields, cross-format mappings, validation and the responsibility of the person approving a localized record.
The IPTC Video Metadata Hub 1.7 Recommendation was approved in October 2025, with a release announcement in November 2025. It includes properties for video content, rights, administration and technical characteristics, with mappings into other standards. The correct implementation must follow the property definitions, repeatability and data types of the version actually exchanged. A human-readable translation can be excellent while the metadata is still wrong; a successful JSON parse cannot tell you whether the translated rights notice grants permission the source never gave.
Start with the question the video record answers
A video record helps someone discover an item, decide whether it is relevant, understand what it contains, identify its provenance, and determine whether it can be used. Different properties answer different questions. The title can help a person recognize the work; a description summarizes its content; a creator field attributes work; a language field can describe the language heard in the video; and rights fields communicate conditions for use. Treating these properties as interchangeable because they all contain words will produce a confused catalogue.
Imagine a documentary clip recorded in Portuguese, described in English and offered for access on a Japanese site. Its displayed title may legitimately have a Japanese translation. The clip is still in Portuguese unless it has another audio track or version. Its creation and rights history do not change just because a Japanese reader encounters it. Every translation decision should begin by asking which real-world fact the property represents and whether that fact has changed.
What Video Metadata Hub actually standardizes
The word “hub” matters. IPTC Video Metadata Hub is a coherent set of metadata properties and mappings that makes it easier to communicate information across systems using different media standards. It is not a new universal video container and does not replace every local storage schema. Its properties can be expressed through technical forms such as XMP and JSON, and the Recommendation provides mappings to multiple established representations including EBUCore and other industry models.
The purpose is semantic alignment. A term such as “creator” should have a documented meaning even when software stores it under different technical names. A translation team must preserve that meaning while changing natural-language values where appropriate. Without that discipline, a crosswalk can give the illusion of success: the destination has a field called Creator and it contains a name, yet the source was referring to an organization that supplied the file rather than the person who created the work. Similar names do not guarantee equivalent roles.
Confirm the Recommendation version first
Standards evolve. IPTC Video Metadata Hub 1.6 introduced or clarified information about rights usage, the main language of a video and supply-chain sources. Version 1.7 added properties relevant to AI-generated content, including AI-system and prompt information. A system using an older property catalogue may not know how to import the newer fields. An implementer who ignores versions may silently discard provenance while presenting the rest of the export as complete.
Record the Recommendation version and the source and destination schema versions in the handoff. Confirm how each system behaves when it sees an unknown property: reject, preserve, ignore or transform. Do not assume that the presence of a JSON field means the receiving application understands its meaning. If a field is unsupported, retain it through an approved extension or sidecar and document the limitation. This is especially important for governance information that can affect how a video is reused or evaluated.
Separate original title, display title and release title
Titles serve discovery, branding and identity. A translated title can make a clip discoverable to a new audience, but it should not overwrite a formally assigned release title or erase the work’s original name. Some organizations use an internal programme title, a public-facing headline and a distributor’s commercial title for the same footage. The translator must recognize which role the source property has before proposing a localized string.
Suppose the internal name is “Storm B-Roll 17” while the published headline is “After the Flood.” Translating the internal label into lyrical prose may make the system harder for a producer to use; translating the public headline may be perfectly appropriate. Preserve established identifiers and version relationships, and label each title according to its intended purpose. A multilingual user interface can display an adapted title without inventing a new asset or implying a new publication history.
Keep descriptions factual and proportionate
A video description may be written for viewers, newsroom staff, archivists or rights officers. It should tell the truth about what the footage shows or says without adding allegations, causal explanations or emotional interpretations that the source cannot support. Localization can change sentence structure, cultural context and vocabulary, but it should not turn a limited observation into a sweeping factual claim. This is especially important for news, health, public safety and historical footage.
For example, a description saying “residents queue outside a clinic after flooding” does not establish that the flood caused every illness in the queue. A natural-language translation should preserve that distinction. Check the video when possible, compare the source caption or editorial script and retain uncertainty markers. A shorter target description can be acceptable, but important qualifications should not disappear simply to fit a character limit. Readability and evidence must be managed together.
Distinguish the main language from metadata language
The language of a video is a property of the recorded or delivered content; the language used to describe it is a property of the catalogue interface or a metadata value. IPTC’s Video Metadata Hub 1.6 change explicitly clarified that its Language property refers to the main language of the video. That clarification matters when a distribution portal offers translated titles but keeps the original audio. Labelling every localized page as a new video language creates false expectations for viewers.
A documentary may be primarily in French with isolated Arabic dialogue and English captions. The record should not describe the whole video as English simply because the metadata has been translated into English. Identify main audio language as required by the property’s semantics and represent additional language information through appropriate supported fields or associated materials. If a newly dubbed version is created, manage it as an actual media version with its own clear relationship. The language tag should not be a marketing shortcut.
Recognize subtitles and transcripts as distinct resources
Subtitles, captions, transcripts and dubbing are related to localization but have different functions. A subtitle track provides timed reading text; captions may also communicate non-speech sounds; a transcript provides a textual rendering of speech; dubbing creates a different performance. A metadata record can describe the presence of these materials, but a translated description is not evidence that the video has become accessible in that language.
For distribution, confirm the availability, completeness and intended purpose of each language resource. A trailer with one English subtitle sample should not be listed as having fully translated English subtitles. If a target locale requires accessibility features, check the delivered tracks rather than trusting a generic metadata label. Record external subtitle and transcript identifiers where appropriate. Careful documentation helps audiences choose the right version and helps delivery teams avoid sending a file that lacks an advertised service.
Preserve names and person identities
Creator, contributor and people-shown information may serve attribution, contextual research or search. Translating names requires restraint. Established personal names, organization names and controlled authority references generally should not be replaced with improvised equivalents. Transliteration may improve reading in another script, but it must not change the identity of the person or organization being described. Where multiple spellings exist, follow an agreed authority system or documented editorial policy.
Imagine a historic interview whose speaker is known under several romanizations. A target-language catalogue can offer a local-script form, but it should link to the same person record. Avoid inferring sensitive characteristics from a face or accent. If the source does not identify a person, do not invent a name to make the metadata seem complete. A good multilingual description makes real people easier to find while retaining the limits of what the archive actually knows.
Creator and supply-chain source are different concepts
The creator of a video is not necessarily the person or system that supplied a file to the current repository. Supply-chain source information concerns how a piece of media entered a workflow or from whom it was obtained. Version 1.6 clarified that such a source may be a system as well as a person or organization. The distinction becomes important when footage passes between news agencies, editing platforms, generative systems and storage services.
A broadcaster might obtain a clip from a syndication feed originally created by a documentary crew. If a translator turns the syndication supplier into the creator, attribution becomes misleading. Keep both relationships distinguishable. A label that is convenient in a user interface should not alter the provenance model. Review the relevant property definition rather than deciding from the English word alone. In chain-of-custody situations, a mistaken role can have practical effects on licensing and verification.
Describe what appears without claiming more
Properties about visible and audible content can support search by people, objects, products, places and events. The challenge is to express what the recording actually supports. A camera shot of an office building does not prove that a particular company owns the building. A product visible in footage does not establish sponsorship. A flag in the background does not necessarily locate the scene. Translation should preserve the scope and confidence of the original assertion.
When a source title is ambiguous, favor an accurate explanatory phrase over a confident but unverified identification. If a person is merely mentioned, do not convert them into someone shown on screen. The difference matters for archives, newsrooms and automated discovery. Connect verified entities to authoritative records where supported, and preserve a plain-language note when the identification is uncertain. The target-language search experience should improve recall without fabricating evidence.
Place names need provenance and context
Geographic information can indicate where footage was made, where an event happened, or the place the subject concerns. These are different relationships. Languages and jurisdictions also have different place-name conventions. An automatic transformation from an old place name to its modern form can erase historical context or introduce political assumptions. Keep the underlying location identity and the source’s temporal context explicit.
A historical recording might use a place name that changed decades later. A localized catalogue may explain the current name as a search aid while preserving the historical wording in the record or provenance note. When geographic coordinates are provided, do not treat them as ordinary words to translate. Ensure any human-readable location label still refers to the same place. If the source record gives only a broad location, resist the temptation to make it precise without evidence.
Dates and publication histories are not synonyms
A video can have a creation date, an original publication date, a modification date and a distribution date. Translating every one as “date released” can create a false chronology. The distinction is important for embargoes, licensing, archival scholarship and newsroom verification. A target-language label should identify what happened on the date, while the underlying date value follows the system’s required representation.
A news clip may have been filmed on Monday, edited on Tuesday and uploaded internationally on Wednesday. All three can be true. If a platform stores them separately, keep that separation through any metadata mapping. Locale-aware display may change the order of day, month and year, but should not change the event or timezone meaning. When the source is uncertain, preserve its uncertainty instead of replacing it with an exact-looking date.
Rights statements describe conditions, not permission inventions
Rights metadata is among the most consequential areas of localization. A sentence that says a broadcaster may request use for educational screenings does not mean commercial redistribution is authorized. A field describing a copyright owner is not a license grant. IPTC Video Metadata Hub includes several rights-related properties; their semantics should be consulted individually rather than handled as one generic free-text note.
For international delivery, distinguish the rights holder, permitted usage, restrictions and any links to authoritative terms. Translate free-text conditions through a reviewer competent in the relevant language and policy context. Retain the source wording and version where needed. Do not broaden territories, extend dates or remove approval requirements to make the result friendlier. If the record is silent on a use, do not infer permission. Metadata should communicate actual rights arrangements, not create arrangements by linguistic accident.
Rights Usage Terms have a specific purpose
Version 1.6 introduced a Rights Usage Terms property describing licensing parameters in free text, aligned with IPTC’s photo metadata vocabulary. That can make production guidance easier to exchange, but it also requires care. Free text is not automatically complete legal advice, and an importer might display it without other contextual restrictions. A localized sentence must be reviewed for consistency with the agreement controlling the video.
Suppose the source says the footage is for use in a single named programme until a particular date. A translation that omits either constraint changes the recipient’s understanding of permitted reuse. Keep reference links, governing agreements and the relevant approval authority available. Use a structured field for a structured fact when the chosen implementation supports one, rather than relying on a long prose sentence to carry everything. If rights information conflicts, escalate it instead of choosing whichever translation seems simplest.
Technical properties are evidence, not adjectives
A video record can describe dimensions, frame rate, duration, encoding and related technical characteristics. These values are not ordinary prose; they describe the actual digital representation or signal. The target language may use different typographic conventions for units, but the machine values must remain accurate and unambiguous. An article or catalogue can say “high definition” for readers, but that phrase must not replace the underlying measurable properties.
When a distributor sends an asset to another region, confirm the receiving system’s technical expectations independently of language. A mismatch in frame-rate interpretation can affect synchronization with subtitles and audio; a mistaken duration can break automated schedules. Do not round numbers merely for style. If a display layer converts units or formats, retain the original values and document conversion rules. A translated label should help the operator understand a property, never quietly substitute an inferred value.
Cross-format mapping needs semantic review
Video Metadata Hub mappings can connect similar properties across XMP, EBUCore, JSON, PBCore and other systems. A mapping is not a claim that every pair of fields has identical behavior in every implementation. Some formats allow repeated values, relationships or localized alternatives that others cannot carry without extension. A crosswalk should record which content is preserved, which transformations are required and where information may be lost.
Consider a repeated series title and episode title in one source and a single title field in a target. Choosing one title may be necessary for a restricted interface, but it is not a lossless conversion. Keep the source record or a structured sidecar. Review the official IPTC mappings rather than aligning fields only by similar spellings. Test actual export and import with values that exercise the difficult cases. Otherwise the success of a simple example can conceal failures in the larger media library.
Understand XMP as an embedding technology
XMP can carry structured metadata associated with media files. Its role differs from the semantic role of the IPTC properties themselves. One describes a way to represent and store metadata; the other defines what the metadata means. Translating an XMP property key is generally wrong. Translating an approved text value or language alternative may be appropriate according to its schema and the embedding application’s capabilities.
Before embedding localized strings, check whether the file format, software and distribution pipeline preserve the required XMP data. Some systems strip embedded metadata during transcoding, while others copy only selected properties. Keep a sidecar or canonical record when necessary, and verify the exported result rather than assuming preservation. Rights and provenance are especially important to check after upload to a platform that may rewrite media files. Successful encoding is not the same thing as successful metadata transport.
Understand JSON as a structured exchange
A JSON implementation of Video Metadata Hub can help connect APIs and catalogue databases. JSON keys, data types, object relationships and arrays are not translatable sentences. A translator should work on eligible string values within the property’s defined semantics. Converting a boolean into a localized word or a machine-readable timestamp into a descriptive date string can cause validation and interpretation errors.
Build a test set with non-Latin text, repeated values, optional fields, null values and special characters. Verify round-trip decoding and validate against the correct schema. Ensure that systems do not silently coerce numbers into strings or replace missing data with guessed defaults. In cross-language workflows, an empty string, an absent property and a genuinely unknown fact can mean different things. Preserve those distinctions so downstream searches and rights decisions do not become unreliable.
EBUCore and IPTC are complementary, not identical
EBUCore is designed around broadcast media metadata and its own semantic and technical structures. Video Metadata Hub provides a useful common set of property meanings and mappings, but it does not erase the differences between standards. A team that has already localized EBUCore may still need to review IPTC-specific language and rights semantics during the interchange. Simply copying the same label into a field with a similar name is insufficient.
A broadcaster might have an EBUCore record with established technical and distribution relationships and wish to expose selected information to a public news portal using Video Metadata Hub. Decide which facts need to travel. Map original values carefully, retain a link to the canonical record and document any information the portal cannot represent. The portal should be useful to readers without pretending to contain the full archive of broadcast engineering metadata.
PBCore and archival descriptions require different care
PBCore distinguishes intellectual audiovisual assets from their material instantiations. Video Metadata Hub can represent an important shared set of content and administrative properties, but an archive may have finer-grained information about original tapes, preservation masters and access copies. A translator mapping PBCore into an IPTC-oriented system must keep that source distinction visible when the destination does not provide an equivalent object model.
Do not take a file-specific technical property from one preservation master and apply it to every version of the audiovisual work. Likewise, a rights restriction on one instantiation might not describe every copy. Where the destination is simpler, select the appropriate representation for the public use case and maintain a robust link to archival detail. Localization is successful when the summary is trustworthy, not when every source value has been forced somewhere into the destination record.
AI provenance deserves its own review
Video Metadata Hub 1.7 added properties for AI-related production information, including the system used, its version and prompt information, where applicable. These properties can help describe how content was made or modified. They must not be treated as decorative marketing text or as proof that a particular process occurred. A source record may contain missing, partial or user-supplied claims. Translation should preserve the distinction between recorded evidence and unverifiable assertion.
If a record says a production used a named AI system, retain the stated system name and version as factual identifiers where appropriate. Translate explanatory prose without inventing capabilities, motivations or provenance details. Prompt information may include confidential, personal or sensitive data, so determine whether it may lawfully and ethically be published before exposing it in another language. A multilingual catalogue should improve transparency while respecting disclosure rules and the limitations of the available evidence.
Do not confuse AI creation with AI-assisted cataloguing
An automated tool may generate a descriptive caption about a conventional camera recording. That does not mean the video itself was generated by AI. The provenance of the audiovisual content and the provenance of its metadata are distinct. Conflating them can mislead viewers and damage trust. Similarly, translating a human-written description with AI assistance does not automatically imply that the footage was synthesized.
Keep separate records of asset production, editorial modifications and metadata-generation workflows where the system allows. If a tool proposes people or locations based on image recognition, review those suggestions before treating them as verified facts. The responsibility for publication remains with the organization. When translating AI-related properties, retain exactly the scope of the source claim and avoid turning a technical process description into a broad statement about authenticity.
Rights and provenance can change through supply chains
A video may pass from creator to agency, agency to broadcaster and broadcaster to platform. At each step, a copy can receive new technical identifiers, distribution restrictions or contextual metadata. This does not mean every version has the same usage rights or provenance wording. Translation must preserve the provenance and rights of the particular object being transferred, while retaining connections to earlier authoritative records.
A newsroom receiving footage from an international partner should know whether it has permission to edit, crop, dub, excerpt or republish the material. A friendly translated rights label is not enough. Verify the formal license and exchange agreement. When the metadata itself does not establish authorization, say so. This protects the newsroom, the original creator and the audience. Accurate metadata helps good editorial decisions; it cannot replace the legal and ethical decisions behind those uses.
Build a field-level change table
A reliable localization project uses a field-level record of changes. For each IPTC property, document its role, the source value, any controlled code, the proposed target value, the reviewer and the check performed. Classify fields as translatable descriptions, controlled terms, identifiers, technical values or legal/provenance information. This avoids sending the entire file to a translation engine without distinguishing human text from structural data.
Include a “not changed” category and explain why. Unchanged fields may be the most important proof of fidelity. If the translation changes a display name, record its approved relationship to the original. If a source value is ambiguous or appears incorrect, raise a separate metadata correction request. Do not silently mix editorial repair with translation. The change table becomes a useful artifact when the same video is later localized into a third or fourth language.
Create a realistic test package
Test with a small set of carefully selected video records rather than starting with the simplest title-description pair. Include a clip with two languages, a historical place name, multiple credited organizations, time-bounded rights, an access restriction, non-Latin titles, two distinct technical renditions and AI-related provenance that may be absent or partial. Export it through the source system, translate permitted values and import it into the destination using the actual production path.
Then check what survived. Did the video language stay the same? Did the right rights statement attach to the right rendition? Did repeated titles remain associated with their roles? Did JSON arrays stay arrays? Were source names and identifiers preserved? Did the destination strip provenance fields it did not understand? A controlled pilot should expose these failures before a library-scale transfer. Document both what passed and what the receiving software cannot express.
Review published display pages, not just source files
The technical metadata can be valid while the website misrepresents it. A page might display the target language as the soundtrack language, hide an embargo condition, drop an accent from a personal name or show an internal title as the public headline. These problems occur after schema validation because rendering templates, search indexes and content management rules make additional decisions.
Open the localized page as an ordinary visitor and inspect the actual presentation. Search for a title using both original and translated terms. Verify that people and places link to the intended records, rights warnings remain visible and accessibility information describes the delivered file. Test a mobile view if that is where users will encounter the catalogue. A successful data exchange is only a midpoint; the user-visible result is part of the product.
A newsroom example with three language markets
Imagine a short current-affairs video made by an independent team and distributed to partners in three markets. Its source metadata includes a title, a factual description, creator and supplier information, a primary spoken language, licensing conditions and technical details. Each partner needs a readable local-language presentation. Begin by identifying the video and its distribution versions. Keep immutable identifiers, rights scope and source information stable while translating the approved public descriptions.
Next map the properties into the partner’s chosen interchange representation. Confirm version support, preserve controlled roles and inspect whether the receiving system displays the source of the footage separately from the creator. If one partner adds a dubbed audio track, treat that as a new content version and update language information on the relevant version, not on the original. Finally check the published pages, download permissions and player selections. The finished result should tell each audience the truth about what they will watch and what use is allowed.
Questions editors and localization teams ask
Is IPTC Video Metadata Hub a video codec or file container?
No. It defines a common set of metadata properties and mappings among technical standards. It can be represented using supported forms such as XMP and JSON. Media compression, packaging and playback are separate concerns. A file with IPTC metadata can still have incompatible video essence, and a playable video can still have missing or misleading metadata.
Can I update the language property when translating a catalogue?
Only when the actual meaning of the property changes. The principal video-language property describes the main language of the video, not simply the language of the web interface. A translated title or abstract does not change the soundtrack. A new dubbed version can require its own legitimate language metadata and version association.
Does a translated rights statement become a new license?
No. The translation should faithfully convey the source conditions and should be reviewed by a person qualified for the context. Rights scope, dates, territory and reuse restrictions must not be invented or broadened. If the source contains an unresolved contradiction, report it instead of choosing the most permissive interpretation.
Are AI-related metadata fields always required?
No. Requirements depend on the applicable property definitions, implementation and workflow. If information is unavailable, do not manufacture a system name or prompt. Preserve legitimate uncertainty, respect confidentiality and verify what the source actually claims. Check the version of the Recommendation supported by both sender and receiver.
Can one field mapping replace every other video standard?
No. Video Metadata Hub is meant to improve alignment across multiple technical and semantic models. More specialized standards may carry relationships, preservation details or technical facts outside a destination system’s scope. An honest crosswalk identifies what has been preserved and what requires another record, extension or sidecar.
The professional standard is semantic fidelity
Good video metadata localization does not make the database prettier at the expense of truth. It helps new readers understand the same footage, rights, identities and history that the source record described. It preserves the main language of the actual media, distinguishes creators from suppliers, keeps technical numbers stable and treats emerging AI provenance as evidence requiring care. With a field-level review, proper crosswalk and end-to-end import test, multilingual metadata can strengthen a media library rather than fragment it.
Useful related readings include EBUCore broadcast metadata, IMF language-version packages and MXF metadata and track identity. For authoritative definitions, consult IPTC Video Metadata Hub, the 1.7 Recommendation, official user guide and IPTC’s November 2025 release notice.