How can you translate an Adobe Captivate course while preserving slide timing, software demonstrations, quizzes, captions and LMS completion? A Captivate project contains more than text paragraphs. It may combine responsive slides, imported screen recordings, trigger-based interactions, audio, subtitles, assessment feedback and project settings. A translated project can look polished while still teaching the wrong screen command or sending incorrect completion results.
This Adobe Captivate localization guide explains how to adapt learner-facing project content, captions, subtitles, software walkthroughs, quiz feedback and published training for multilingual audiences. It distinguishes current Adobe Captivate workflows from Adobe Captivate Classic where their menus and file formats differ. Readers will learn how to plan language copies, work with exported DOC content, validate SRT or VTT subtitle timing, handle image-based UI text, and test output on the devices and LMS platforms where learners actually study.
The essential starting rule is to identify which Captivate product and project version you have. Adobe’s documentation describes export and import of project content and closed captions as Word DOC files for supported workflows, while current Captivate also documents SRT and VTT captions. Classic documentation additionally describes XML export of project content. These capabilities are not a promise that every project type or version supports the same import route. Test the exact round trip on a disposable copy before scaling up.
The answer in one minute
The practical rule for Adobe Captivate is to preserve meaning, interaction and evidence together. First identify the original course revision and all learner-facing elements. Next select a translation path that your installed product actually supports. Then review the language in context, test the same learner actions, and publish only an accepted target edition. A translated word count or successful file import proves very little unless the finished course still works for the person who studies it.
The strongest quality question is simple: can a learner who does not understand the source language read the material, choose the correct action, understand a mistake and finish the course with the expected record? Ask that question of every major page, assessment and resource. It converts localization from a cosmetic task into an observable educational outcome.
What must stay true when a course changes language
Captivate localization is a controlled adaptation of an executable lesson. The learner reads an instruction, waits for a cue, selects a screen object and receives feedback. The written text, visual object, timestamp and branching logic jointly create that learning event. If a translated button label points at the wrong place on a recorded interface, the lesson cannot teach the intended skill even if the grammar is immaculate.
Separate authored text from captured pixels. Text objects and supported caption exports may be translated through editable project content, but a screenshot contains words as part of an image. If the depicted application is available in several languages, it is often more reliable to capture the target-language application again rather than place a translation on top of an English interface.
Make a language edition from a source revision. Record the application version shown, Captivate product version, slide count, project properties, quiz rules, audio assets and intended LMS standard. Preserve an untouched original, and test the project with the original application workflow before editing. A translation cannot rescue an unstable source tutorial.
Build a glossary around user actions. Distinguish Save, Submit, Publish, Send and Finish, and capture which action exists in the actual application. On-screen command names should match the localized software interface where possible. Keep feature names and menu paths exact for the audience’s installed version, not for a hypothetical future version.
Finish by testing behavior, not merely import. Reopen translated slides, inspect object geometry, validate timing, deliberately choose incorrect answers, run navigation with a keyboard and publish an LMS test package. In a simulation, a misplaced click region is as serious as a mistranslated instruction, and a misleading status message can be worse.
A complete worked example
Imagine a Captivate onboarding module teaching employees to request access to a shared system. The project begins with a recorded login, a text caption describing the request form, a step-by-step simulation, a short quiz and a completion message. The organization needs a German edition. Before translating, note the precise web application’s version, the audience’s regional interface language and the approval process for requesting access.
An English caption reads ‘Choose Request access, then select your role.’ The German application may display different button wording and capitalization. The translator should not invent a menu label from a dictionary. Instead open the actual localized interface, record the displayed string, confirm the order of clicks and check whether the next screen uses the same role names. This is an example of translation as interface verification.
Prepare the project-content export according to the specific Captivate version. Work in a copy, retain the file structure expected by import and translate only the designated human-facing fields. If the project has voice narration, also prepare a target-language audio plan. Subtitles may be imported separately from SRT or VTT files, but their cue durations must remain within the associated audio duration and must not overlap improperly.
Rebuild or adapt the affected screen capture, align captions to the new timing, and complete the simulation as a learner. Then attempt an incorrect click, use the keyboard where supported, and test the quiz immediately below and above its passing threshold. Compare LMS completion and score records across language editions. The result is acceptable only when the learner can follow correct actions on the target interface without guessing.
30 practical Adobe Captivate localization situations
The situations below are deliberately concrete. Each describes a real decision a course team might face, the likely failure, a safer approach, and a test that a reviewer can perform. The examples are applicable across regions; actual regulatory requirements and software feature availability must be verified for the intended audience.
1. Confirm the project generation
Picture the actual learner encounter: A team inherits Captivate Classic source files and newer Captivate projects in the same training catalog. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Assuming their editing menus and export formats are identical can make a planned localization impossible. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Identify the exact application and project version, consult documentation for that version and run a one-slide export/import pilot. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Confirm that the pilot preserves caption text, object properties and all intended project behavior. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
2. Copy before translation
Consider a common handoff: A language coordinator opens the only editable version of a compliance course. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Replacing source strings in place may destroy the approved original and make later updates untraceable. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Create a protected source snapshot and a named target-language project with version and ownership recorded. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Open both projects after the first import and confirm that the source has not changed. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
3. Export DOC content
At this point in a multilingual release, The supported Captivate workflow exports project content and closed captions into a Word document. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Editing IDs or nontranslation columns can disconnect the replacement text from its original objects. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Preserve structural columns and edit designated localized-content fields using the original exported DOC document. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Import into the test copy and inspect the content of several nonadjacent slides. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
4. Respect document provenance
A typical course-maintenance problem looks like this: Two developers export caption files from different versions of the same course. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Importing a file against a different project state may produce mismatched content or unexpected failures. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Keep each export bound to its exact project revision and do not merge files without a controlled comparison. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Match exported slide and item identifiers to the intended project before any production import. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
5. Check Captivate Classic XML
Picture the actual learner encounter: A legacy simulation stores dynamic objects and project properties that are not all exposed as ordinary slide captions. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Translation teams may overgeneralize from a DOC-caption workflow and overlook dynamic content. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: For Classic projects, assess the documented XML content-export route and protect the XML structure during any supported editing. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Import into a duplicate and compare object text, properties and quiz behavior. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
6. Translate button captions
Consider a common handoff: A captured form uses Continue, Confirm and Submit as three separate steps. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Using one translated word for all three can erase the difference between navigating, accepting and sending data. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Document the application action behind each button and align its wording with the actual target-language UI. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Ask a target-language learner to predict the consequence of each button before clicking. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
7. Rerecord localized screens
At this point in a multilingual release, A software demonstration shows menus in English while the target learner uses a French user interface. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Subtitles alone cannot resolve differences in menu position, wording or visible fields. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Record the same approved workflow on the localized application and retain a map of corresponding steps. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Follow the final footage on a real target-language account and verify the click path. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
8. Preserve click areas
A typical course-maintenance problem looks like this: A translated caption is added beside a hotspot used in a software simulation. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Changing the visual arrangement without checking hit areas may make a correct click appear wrong. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Review hotspots, bounding boxes and responsive repositioning after localization changes. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Complete the interaction with both mouse and keyboard if the interaction supports them. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
9. Verify drag and drop
Picture the actual learner encounter: A quiz asks learners to drag task names onto departments. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Translated labels may not fit the draggable objects or may subtly change which department owns a task. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Review both the classification logic and the visible label geometry with a subject expert. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Drag every item to correct and incorrect targets and confirm the same feedback and scoring. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
10. Localize success messages
Consider a common handoff: A simulation displays ‘Request submitted’ only after the final form is sent. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: A mistranslated message may suggest approval is already granted, which changes the employee’s expectations. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Translate the actual system state rather than a tempting synonym and align it with the workflow status. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Compare displayed confirmation with the application audit log after a successful request. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
11. Preserve warning severity
At this point in a multilingual release, A safety module uses a high-priority warning before a hazardous operation. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Softening ‘must not’ into ‘should avoid’ changes the permitted behavior. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Classify and review safety-critical language separately with the appropriate domain authority. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Present the warning to bilingual users and ask them to explain the prohibited action. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
12. Align closed captions to audio
A typical course-maintenance problem looks like this: Narration explains a process while captions appear over the demonstration. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: A correct translation can be unreadable if its cue begins too late or remains after the corresponding action. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Time the caption file against the actual target narration and use natural reading units. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Watch the lesson muted and confirm that each displayed cue matches the visible stage. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
13. Validate SRT timestamps
Picture the actual learner encounter: A subtitle file is imported in SRT format for a slide with spoken audio. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: A zero-length cue, overlap or end beyond audio duration can cause an import error or unusable subtitles. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Validate chronological order, positive cue durations and time limits before importing the supported format. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Re-import and play the full slide to check the last subtitle and any overlap. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
14. Validate VTT captions
Consider a common handoff: A caption editor provides WebVTT files with styling and timing cues. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Reusing a file with incompatible syntax or unsupported font choices can alter the caption appearance. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Keep WebVTT syntax valid and check how Captivate handles styling and fallback fonts. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Preview captions in the actual published runtime, not only in a text editor. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
15. Replace a bad caption file
At this point in a multilingual release, A first translation import contains inaccurate names for three commands. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Patching each displayed cue manually without a source record makes future correction harder. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Update the approved caption file, use the documented replace route and verify the whole affected slide. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Compare the new output against the corrected subtitle list at every timestamp. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
16. Control slide timing
A typical course-maintenance problem looks like this: A translated instruction is longer than the original and appears beside a moving pointer. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: The learner may still be reading when the next action begins. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Increase display time or redesign the cue sequence within project constraints, keeping the demonstration’s causal order. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Have a reader unfamiliar with the course follow the instructions at ordinary speed. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
17. Review responsive slides
Picture the actual learner encounter: A paragraph that fits a desktop layout wraps into an overlapping card on mobile. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: A clean editor view may conceal cut-off text in the published responsive output. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Test the supported viewport sizes and simplify wording without dropping essential conditions. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Read the complete page on a small screen and tap every interactive control. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
18. Translate quiz stems
Consider a common handoff: An assessment asks which action must be performed before requesting access. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Moving a time qualifier or negation can reverse the correct answer. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Review the stem, alternatives, feedback and answer key as one assessable unit. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Attempt the item in both languages and confirm equivalent difficulty and scoring. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
19. Protect numeric thresholds
At this point in a multilingual release, A course requires an eighty-percent pass score and a limited number of attempts. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: A localized explanation may promise unlimited attempts or treat the percentage as a grade earned. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Keep course rules as verified settings and phrase their meaning precisely in each edition. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Test scores on both sides of the pass mark and repeat an attempt to check the actual limit. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
20. Translate feedback states
A typical course-maintenance problem looks like this: A learner who selects an incorrect option sees advice explaining the next step. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Generic encouragement can hide a safety correction or cause a learner to repeat the error. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Preserve the reason and corrective action, not merely the positive tone of the original feedback. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Intentionally make every common error and review the response it triggers. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
21. Review variable placeholders
Picture the actual learner encounter: A confirmation slide inserts the learner’s name into a sentence. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Translating a variable identifier or ignoring target-language word order can display nonsense. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Protect tokens and plan grammar around the values that the runtime supplies. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Test sample names of different lengths and inspect the rendered sentence. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
22. Check conditional actions
Consider a common handoff: A branching slide shows different content depending on an earlier answer. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Reworded labels can accidentally imply a choice that triggers a different destination. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Maintain a decision table of triggers, conditions and target screens alongside translated copy. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Execute every branch and compare the instructional consequence with the source. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
23. Handle voice narration
At this point in a multilingual release, The source video has synchronized spoken instructions and animated highlights. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Replacing the speech without retiming the motion can draw attention to the wrong object. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Record with a pronunciation guide, measure actual target audio and adjust slide or animation timing as necessary. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Review the exported lesson with target audio and no English subtitle aid. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
24. Plan text-to-speech pronunciation
A typical course-maintenance problem looks like this: A project includes product names, abbreviations and person names in synthesized speech. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: A voice may pronounce technical tokens letter by letter or stress them misleadingly. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Prepare a pronunciation lexicon and listen to the rendered audio rather than assuming written spelling controls speech. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Ask a speaker of the target language to identify each name as spoken. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
25. Revisit screen-reader labels
Picture the actual learner encounter: An image button is visually obvious but read aloud as an untranslated filename. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Sighted review may pass while assistive technology exposes a broken or ambiguous task. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Provide meaningful accessible labels and check focus and reading order within the chosen Captivate version. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Navigate the translated interaction without sight and identify each actionable control. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
26. Adapt imagery with text
Consider a common handoff: A process map contains embedded English labels that are part of a graphic. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Translating a nearby caption cannot make the diagram fully understandable to a target-language learner. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Obtain an editable illustration or recreate the image with translated labels and equivalent visual relationships. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Ask someone to explain the diagram without reading its surrounding paragraph. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
27. Handle dates and symbols
At this point in a multilingual release, A technical quiz uses decimal numbers, units and region-sensitive date notation. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: An attractive local format can introduce an ambiguous value or change the mathematical answer. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Specify units, validate numeric parsing and avoid ambiguous date formats in high-stakes instructions. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Have reviewers independently calculate the required value from the target-language screen. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
28. Protect external destinations
A typical course-maintenance problem looks like this: A lesson links to a live application, shared policy and help desk form. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Translated link text may promise a localized destination that the organization has not provided. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Verify destinations and explain when a linked service remains in another language. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Follow every link under a learner account and confirm that permissions and language expectations match. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
29. Package for an LMS
Picture the actual learner encounter: A translated Captivate edition is published to a learning platform as a tracked course. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: A correct preview does not establish successful score and completion reporting. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Test the supported publishing standard and project settings in a staging LMS with separate language enrollments. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Compare completion, score, resume and failure behavior against the approved source. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
30. Track change history
Consider a common handoff: An application update renames a menu after the localized Captivate project has been released. This matters because the learner sees a functioning training experience, not a collection of strings waiting to be checked. The relevant meaning includes the action that the text requests, the surrounding visual cue and the consequence of making a decision. Before changing wording, identify who owns that meaning and how it will be demonstrated in the finished course.
The weak approach is easy to understand but risky: Learners may follow obsolete screenshots even while accompanying text is technically correct. Even a fluent sentence can conceal a wrong state, a misplaced control or an assessment whose difficulty has shifted. Compare the source and target in their full task context rather than approving the translated phrase in isolation. Where rules or user permissions are involved, document them as requirements instead of hoping the localizer can infer them.
A stronger method is more deliberate: Record application and project revisions, update affected media and retire misleading older packages. Give reviewers enough context to judge whether the intended instruction survives, and preserve technical data that is not meant to change. Distinguish a legitimate regional adaptation from an accidental change to an objective, score, workflow or identifier. If stakeholders disagree about a term, settle the underlying action first; the vocabulary usually becomes clearer after that.
The acceptance test is practical: Reperform the exact workflow against the current target-language application before republishing. Repeat that test with the permissions, device size and language settings a real learner will have. Make a note of unexpected behavior, correct it at its source, and rerun the affected path. The work is not complete because an editor saved a file: it is complete when the experience is understandable and produces the same justified educational result.
Four-week practical localization plan
Week 1 — Define the source and the risk
Inventory all course pages, actions, quizzes, media, labels and external resources. Select the exact source revision, target locales and reviewers. Distinguish ordinary learning copy from safety-critical or legally consequential wording. Create glossary decisions for ambiguous action verbs and a list of values that must not be translated, including identifiers and scoring settings. Select a small but representative lesson for a trial translation. At the end of the week, the team should know what is in scope, what software version is being used and what constitutes acceptance.
Week 2 — Translate with context
Produce a first target-language edition through the supported workflow. Review text in the actual learner view, not only in an exported file. Translate quiz items together with their answer keys and feedback; inspect hidden layers and dependent media. Resolve terminology disagreements against learning outcomes and real workplace or classroom actions. Make local substitutions only where approved. At the end of the week, the team should have a readable complete draft, an issue log and a clear list of media or technical gaps.
Week 3 — Validate learner behavior
Run the course on desktop and narrow mobile screens, check captions and keyboard access, and explore every important branch and quiz outcome. Use target-language reviewers who have not memorized the source lesson. Test links, labels, graphics, instructions and any LMS integration in an appropriate staging environment. Compare pass, fail, incomplete and retry behavior with the source edition. At the end of the week, issues should be graded by impact and blocking defects should have been corrected and retested.
Week 4 — Publish an auditable edition
Review the final package or managed language variation against the acceptance record. Verify that only approved editions are visible to learners, and document their revision, locale, release owner and date. Record exactly where the course is assigned or linked, including LMS references and region-specific resources. Provide a simple route for reporting a confusing sentence or broken interaction. Schedule maintenance triggers for future policy, application or source-course changes. A language release is a maintained educational product, not a one-time export.
Frequently asked questions
Does Captivate use the same localization steps in every version?
No. Captivate and Captivate Classic documentation describe differing features and menus. Identify the exact product generation first and verify export/import support with a test copy.
Can I use Word DOC files for localization?
Supported Captivate workflows document export and import of project content and captions via DOC. Preserve the exported structure and edit only the intended translation fields.
Does translating captions translate recorded software screens?
No. Words baked into captured screens remain part of the image or video. Re-record the localized application where accuracy requires it.
Can Captivate import SRT and VTT?
Adobe documents SRT and VTT caption workflows for current Captivate. Check valid timing, syntax and slide audio before importing.
What should remain unchanged during translation?
Protect quiz keys, passing thresholds, project identifiers, variables, navigation actions and completion rules unless an approved change is separately required.
How do I test subtitle timing?
Compare actual audio duration with cue start and end times, check for overlap and read the published result at ordinary speed.
Is an editor preview enough?
No. Test the published course on expected devices and in the target LMS, including incorrect answers, incomplete exits and passed attempts.
How often should localized simulations be updated?
Whenever the recorded application or training policy changes materially. Tie each language release to the application and source-project revision used.
Official documentation and related reading
Product features, licensing and file-format support can change. Check the vendor’s current documentation against the exact installed or subscribed version before adopting any procedure. The official sources below support the platform-specific guidance; the eduKateSG links place this guide inside the broader translation and education-interoperability curriculum.
- Adobe — Closed captions in Captivate
- Adobe — Captivate Classic localization
- Adobe — Captivate Classic XML export
- Adobe — Current Captivate user guide
Continue learning within the translation architecture
- The education and learning translation system
- SCORM 2004 localization
- xAPI localization
- LMS and course-level localization
A useful final question for every Adobe Captivate project is not ‘Did we translate every sentence?’ but ‘Can the target-language learner perform the same meaningful task, receive fair feedback, and rely on the recorded result?’ When the answer is supported by evidence from the released course, the language work has become education that travels.