H5P translation is more than changing the language of a quiz question. A tutor can translate every sentence in an interactive activity and still damage its educational purpose if a correct-answer flag moves, a distractor becomes plausible, a fill-in-the-blank no longer matches accepted answers, or feedback gives away the solution before a learner responds. Reliable H5P interactive content localization QA verifies the learning objective, the meaning of each prompt and the behavior of the completed activity in the target language.
People searching how to translate H5P content, H5P language switcher, translate H5P interactive video, H5P multiple choice translation, H5P text overrides and translations, or H5P quiz localization are often trying to prepare an existing learning activity for another class. This guide owns that one reader job: adapt the learner-facing H5P activity while protecting its answer logic and performance evidence. It does not replace an article about translating the Moodle interface, building a Storyline 360 package or localizing an entire learning management system.
Begin with one original activity and record what it assesses. Duplicate or reuse it through an available, authorized H5P workflow; translate one prompt and its answer options; keep scoring and branching identities intact; then complete both the source and target versions as a learner. If both versions test the same skill and award results under the intended rules, the translation has passed a meaningful first test. A language dropdown alone cannot prove that.
Quick answer: preserve the educational contract
Every interactive exercise contains an instructional contract: the learner sees a prompt, interprets it, chooses or enters an answer, receives feedback, and may receive a score or completion state. Translation should alter the natural language but preserve what a correct response demonstrates. A target question becomes invalid if a clue disappears, a negative changes scope, a new grammatical ambiguity appears, or the response-control logic remains tied to source-language words that are no longer visible.
Use four checks. Meaning: does the target prompt ask for the same knowledge or action? Answer: are the designated correct and incorrect responses still valid? Interaction: does the learner see the right feedback, navigation and timing? Evidence: do scores, completion and learner records still mean what the educator expects? These checks apply differently to Multiple Choice, Fill in the Blanks, Drag and Drop, Interactive Video and other H5P content types.
Do not treat “translate the H5P interface” as “translate the lesson.” Built-in buttons such as Retry and Check can be controlled through language or text-override options in supported editors, while author-written questions, answers, narratives and media still need content translation and review. Confusing those two layers is one of the commonest reasons an activity is labelled multilingual while its actual lesson remains in the source language.
1. Decide what a translated activity must prove
Before exporting anything, write the learning objective in plain language. For a Mathematics exercise, it might be “The pupil identifies equivalent fractions without using a calculator.” For a Science quiz, it might be “The pupil distinguishes evaporation from condensation using an everyday scenario.” For English vocabulary, it might be “The pupil selects a word that fits the emotional tone of a sentence.” Each objective gives the translator a stable target even when the grammar must change significantly.
A translation is not automatically valid simply because every source word has an equivalent in a bilingual dictionary. If the source language uses a pun, a cultural reference, or a grammar feature the target does not share, literal translation may change what the child is being tested on. The author and target-language educator should decide whether to adapt the item, replace it with an equivalent example or exclude it from cross-language score comparisons.
Record the item ID, content type, age level, objective, source revision, scoring rule and any curricular references. A concise inventory prevents the team from losing track of which interactive object the translated words belong to. For a larger lesson, the objective belongs to each assessed item, not only to the overall course title.
2. Separate H5P’s player language from its authored content
H5P’s language switcher and text-override features can change many built-in interface phrases. That is useful for buttons, result messages and navigation. It does not guarantee that a teacher-authored question or an embedded video changes language. A Multiple Choice activity might display a French “Check” button while its question and explanations remain English because the author has not translated the content fields.
List the text by owner. Player interface includes built-in labels supplied by the content type or player. Author content includes prompts, answer choices, hints, scoring explanations, captions and learning narratives entered by the tutor. Embedded media includes image text, audio speech and video subtitles. Hosting platform includes the LMS page title, navigation and account controls. Each layer may have a different editing workflow and approval responsibility.
Test each layer separately before declaring the target activity complete. Change the interface language to verify built-in buttons, then inspect authored fields, then complete an interaction. If the interface changes but the quiz does not, the issue is not necessarily an H5P bug: those two kinds of language are stored and maintained differently.
3. Clone the activity before changing its language
A translated version should usually begin as a controlled copy of the approved source rather than editing the only existing activity in place. H5P provides reuse, copy and download workflows where the host and author permissions allow them. Its official documentation explains that a .h5p file may be downloaded when the Reuse option is enabled and can be uploaded into another compatible H5P authoring environment. Availability depends on platform and organization settings.
Copying protects the source and provides a reference for scoring comparisons. Give the new activity an unambiguous locale-aware working name while retaining a link to its source revision. Do not casually copy a production assessment and overwrite it with another language if existing student reports or links rely on the old activity. Ownership, learner records and content identity may be platform-specific; verify the host’s behavior before replacing anything.
After the copy, compare the number and sequence of interactions, expected maximum score, media files and feedback settings. A reuse workflow may preserve content faithfully, but the QA owner should still verify it. A source-to-target inventory is especially valuable for nested Interactive Books or presentations where one visible activity contains several child activities.
4. Respect export and reuse limitations
An .h5p package is a transport format for H5P content; it is not a guarantee that every hosted environment can import every content type or version. Compatible libraries, hosting permissions and feature support matter. Official H5P help notes that download, upload and copy/paste depend on enabled reuse functionality, and its documented user workflow does not offer universal bulk download or upload. Do not promise automatic mass migration from a few successful manual tests.
When a package is obtained legitimately, keep the original intact. Use a copy for inspection or authorized transformation. Do not edit internal package files by search-and-replace unless the team understands the relevant content-type schema and has a validation workflow. JSON keys, interaction identifiers, media references and scoring options may look like text but represent application structure. A translator should not be expected to distinguish them from prose in an undifferentiated export.
For a scalable project, build a manifest listing source activity, target locale, host, content-type version, media dependencies and reviewer. A manual duplication process can still be rigorous when it is small and documented. For many activities, a scripted workflow is worthwhile only after a no-change round trip proves package compatibility and data preservation in the target host.
5. Translate prompts as complete instructional acts
A question prompt usually does more than describe a topic. It tells the learner exactly what kind of response counts. “Choose all correct statements,” “Choose the best statement,” and “Choose one incorrect statement” are materially different instructions. The target translation must preserve the requested action and the scope of qualifiers such as all, only, except, not and most likely. Small words can change which option is correct.
For example, a Science prompt may ask which substance “does not dissolve” in water. If the negative particle is lost in translation, the existing correct-answer flag becomes wrong even though every option is well translated. In a Mathematics word problem, “at least” and “more than” produce different inequalities. The target reviewer should solve the translated question without looking at the source answer key before checking whether the result agrees.
Use a learner-oriented reading test. Read the target prompt, hide the answer controls and ask what action a child would perform next. If the reviewer cannot describe a single clear operation, revise the instruction. This catches ambiguity earlier than an end-of-lesson score report.
6. Preserve answer identity rather than answer position
Many H5P question types store correct-answer settings alongside the options they evaluate. When translating a Multiple Choice item, the target text should change while the logical identity of each option remains stable. Sorting translated answers alphabetically without preserving which option is marked correct can break the answer key. Likewise, replacing a poor distractor with an entirely different concept requires educator approval, not just translator discretion.
Create a mapping table for each assessed item: source option ID or order, source meaning, target meaning, correct/incorrect status and explanation. A reviewer should verify it after editing. For content types where choices can be randomized at runtime, test the randomized state as well as the editor order. A simple static screenshot may hide a mismatch introduced by programmatic reshuffling.
Ask the target-language educator to answer the translated question independently. Compare the selected answer with the configured score. If a fluent target option becomes a second defensible answer, the assessment has changed even if its stored correct flag did not. The right fix may be a more precise distractor rather than altering the original learning objective.
7. Check distractors for accidental clues
A good distractor is plausible enough to reveal a misconception but not so ambiguous that a knowledgeable learner could reasonably choose it. Translation can weaken this distinction. English options may be similar in length, while the target correct answer becomes conspicuously longer. One option may contain an unnatural borrowing, or a distractor may use terminology that excludes it immediately. The result tests language clues instead of the intended subject knowledge.
In a fractions quiz, “one half,” “two quarters” and “three eighths” may serve a specific reasoning task. Translate the concepts while keeping their mathematical relationships. In a vocabulary quiz, a distractor chosen because of an English sound similarity may not work at all in Malay or Mandarin. If the pedagogical mechanism depends on source-language form, a target-language replacement exercise may be necessary.
Have the educator explain why each incorrect choice is wrong. If that explanation changes after translation, classify it as an item adaptation. Record the adaptation so later score comparisons do not assume perfect equivalence. Quality sometimes requires changing surface details; it never permits changing assessment meaning invisibly.
8. Treat feedback as part of instruction
Feedback does not merely say whether a response was scored correct. It may teach a misconception, model a solution or encourage another attempt. If the answer choices are translated but the explanation remains English, the learning journey breaks at the moment a struggling pupil needs support most. Translate positive feedback, wrong-answer feedback, hints and final summaries, and verify which conditions trigger each one.
Check sequencing and disclosure. A hint should not reveal the answer outright when the exercise expects the learner to reason; a wrong-answer explanation should not contradict the corrected option; and a general success message should not claim mastery after a partial score. A translation that turns “Try comparing the denominators” into “Compare the numerators” can change the method being taught.
Perform both a deliberately correct and a deliberately wrong attempt. Observe the exact feedback and next action in each language. Then use a partially correct response when the content type supports it. This tests whether feedback is connected to the learner’s state, not just whether its words sound encouraging.
9. Test partial credit and multiple correct answers
Some interactive activities permit multiple selections or partial credit. Translation may change how students interpret whether they should choose one or several answers. A target prompt that says “Select the correct answer” when two options are intended creates a false failure for a careful learner. Similarly, a translation that changes a distractor into a correct statement can make the scoring rule unfair.
Document the correct-answer set and expected score for every representative combination. The educator should be able to explain why selecting just one of two correct responses receives a particular score under the configured behavior. Do not infer the policy from a single attempt, because question-type settings can vary. Test deliberately complete, incomplete and incorrect selections.
A translated activity can technically produce the same maximum score as the source while no longer measuring the same knowledge. Score-parity testing therefore needs semantic review. A correct-answer flag is data; educational validity requires the target question still warrants that flag.
10. Translate Fill in the Blanks with target-language grammar in mind
A cloze question or Fill in the Blanks activity can be particularly sensitive to language form. If the source expects the word “run,” the target language may require an inflected form depending on person, tense or agreement. A direct dictionary equivalent inserted into the original sentence may no longer fit. The answer acceptance settings must match the new target sentence, not the source spelling.
Rewrite the full sentence naturally before defining acceptable answers. Then consider legitimate alternate spellings, diacritics, spacing conventions and grammatical variants. Overly strict acceptance can punish correct pupils; overly broad alternatives can make an incorrect answer pass. The tutor should decide which variations demonstrate the intended skill and which do not. Test each accepted alternative and at least two near misses.
This is also a warning against simply translating the answer list while leaving the surrounding sentence structure fixed. A target language may require changing word order or the placement of the blank. The stable element is the learning goal, not the English character position where a missing word once appeared.
11. Keep drag-and-drop targets meaningful
Drag-and-drop tasks often combine written labels with spatial target areas. Translating the labels can make them longer and cause overlap with drop zones or neighboring choices. It can also alter the conceptual relationship between items: a term that was a clear category label in English may become ambiguous in another language, or an image may embed source-language text that has not been translated.
Inspect target geometry after changing words. Confirm that every draggable item is readable, every drop area is visually identifiable, and the correct target mapping remains intact. Do not swap target identities simply because the translated ordering looks prettier. For younger students, color and proximity can create accidental hints, so test whether the learner can solve the activity by understanding concepts rather than layout clues.
Run the activity with mouse, touch and keyboard where the content type and host support those interactions. A target label that looks fine on desktop may obscure the drop zone on a tablet used in school. Accessibility and scoring reliability are part of localization quality, not optional decoration.
12. Translate Interactive Video as synchronized learning content
An H5P Interactive Video may contain a video track, captions, timed text boxes, quiz overlays, bookmarks and links. Translating the base subtitles is only one layer. If the spoken video is English but the overlay asks a French question, the teacher must decide whether that question is answerable from French captions and whether pupils have enough time to read it. The translated overlay can require more space or reading time than the source.
H5P’s official Interactive Video guidance describes adding interactions at selected timestamps and setting visibility periods, including whether playback pauses. Preserve the connection between the event in the video and the question about it. If a translated caption runs later than the original sentence, the overlay may appear before the relevant fact is readable. That is an instructional-timing problem, not a spelling problem.
Test from a cold playback at normal speed. Watch the relevant scene, read captions, interact with the question and resume the video. Verify that the response refers to the event just shown and that no overlay conceals important caption text. A still image from the editor cannot demonstrate synchronization.
13. Review video captions, narration and interaction text together
Captions should describe the spoken content accurately, while on-screen interactions may paraphrase or assess it. They do not need identical wording, but they must agree on facts, names, quantities and causal relationships. If a video narrator says “steam cools and forms droplets” but the target quiz calls the process evaporation, the learner receives contradictory evidence. The error may lie in the caption, quiz or original instructional plan.
Maintain a small terminology sheet for scientific terms, historical names, mathematical concepts and recurring character names. The sheet should document concepts rather than force unnatural identical sentence structures. For example, a technical term can have one preferred translation while verbs and surrounding explanation vary naturally with grammar.
When dubbing is not available, check whether the intended audience can realistically follow original audio with translated captions. A Primary student may need more reading time than an adult. If the source video moves rapidly, pausing at an interaction may help; alternatively, the teacher may need an adapted video, separate transcript or slower-paced activity. Translation cannot repair impossible reading conditions through word choice alone.
14. Inspect images containing source-language instructions
Images may carry labels, legends, diagrams, chart annotations or captions embedded in pixels. A lesson about the water cycle might translate the text below a picture while leaving “condensation” and “evaporation” in the diagram itself. A pupil who relies on the diagram receives a mixed-language task, and a screen reader may not have access to the labels at all. Inventory image text before calling the activity finished.
Where rights and source assets permit, replace embedded text with editable labels or provide a properly localized version of the image. Preserve numerical data, diagram arrows, relationships and units. An editor who redraws a graph solely to translate its title must not move a plotted point or reverse a direction arrow. The diagram carries evidence, not just decorative words.
Check image alternatives and surrounding explanations. Alt text should describe the relevant learning information rather than merely say “picture.” If an image’s language changes but its alt text remains English, the activity is inaccessible to some learners. Provide equivalent information suited to the target audience and verify it in the final host environment.
15. Protect numerical facts and units
Mathematics and Science activities are especially vulnerable when text and data are mixed. Translators may need to localize decimal punctuation, unit names or wording around a quantity, but the quantity itself and the required operation must stay true. “Three quarters of a litre” cannot become “three litres,” and a target that changes an instruction from divide to multiply alters the assessed skill.
Treat numbers, formulas and unit conversions as controlled content. If a source example uses 2.5 metres and a target audience uses a decimal comma, the display may change while the underlying value remains 2.5. If a lesson genuinely requires adapting units, that is a pedagogical decision and may necessitate adjusting diagrams, answer keys and scoring tolerances together. It should be tracked as adaptation, not hidden inside translation.
Solve the target problem from its own text. A tutor who reads the source answer first may miss a numerical contradiction introduced in translation. Comparing independent solutions provides stronger evidence that the assessed mathematical relation survived.
16. Be explicit about age, curriculum and local context
A translated activity written for one school system may refer to grade labels, examinations or cultural objects unfamiliar to another audience. A word problem about a familiar bus route can be adapted to another transport context if the numbers and relationships remain equivalent. But translating “Primary 6” into an apparently similar foreign grade may misstate age or curriculum scope. The tutor must distinguish literal terms from true educational equivalence.
Record the intended age range and prerequisite knowledge for the translated version. If the source item assumes knowledge of a local holiday, legal rule or school practice, a target-language note or replacement example may be necessary. The adaptation must not give away the answer or remove the reasoning challenge. Cultural relevance is a support for comprehension, not permission to change the test construct.
A fair comparison asks whether a competent learner with the same subject knowledge—but not the source culture—can understand and solve the target item. If failure depends on unfamiliar cultural trivia rather than the intended concept, revise the item or classify the assessment as not directly comparable across locales.
17. Keep question order and navigation logic stable where needed
A translated activity may contain branches: after an answer, the learner moves to feedback, another question or a remedial explanation. Question order can matter because one item teaches a concept used in the next. Reordering for grammatical convenience can undermine the learning progression. If the author changes the sequence deliberately, the QA record should show that it is a pedagogical adaptation requiring renewed testing.
For nested lessons, create a map of sections, questions, branches and end states before copying. Mark which nodes require language changes and which links or IDs should remain stable. After translation, walk every reachable path with representative answers. A successful straight-through correct path does not prove the remedial or partial-credit branch works.
Verify labels on navigation controls and end-of-activity summaries. A source activity may say “Next question,” whereas the target translates it as “Finish” on an intermediate step. The navigation action still advances correctly, but the learner receives a false instruction. Rendering and behavior must agree.
18. Check learner accessibility in the target language
An activity is not accessible merely because its main paragraph has been translated. Learners may rely on keyboard navigation, screen-reader announcements, captions, clear focus order, sufficient reading time or high-contrast layouts. A long translated label can become clipped at large text sizes; a right-to-left script may require different spatial presentation; a complex question might require clearer sentence segmentation without lowering conceptual difficulty.
Review the content type’s accessibility capabilities in the installed H5P version and host. Not every interaction provides identical support, and translation alone cannot add a missing keyboard behavior. Where a particular content type creates a barrier for the intended learner, consider an accessible alternative activity that tests the same objective. The goal is equivalent opportunity to demonstrate knowledge, not forced dependence on one widget.
Test the target activity with the available assistive tools, not just by reading the text aloud yourself. Check that screen-reader output includes the actual question, answer options, response feedback and completion state in a sensible order. When video is involved, verify captions and whether overlays obscure information needed to answer.
19. Preserve the interpretation of scores and completion
Scoring is evidence about learning, not merely a counter on the screen. After translation, a maximum score of ten should still represent ten comparable achievements if the activity is intended to support cross-language comparison. If one target version contains an easier distractor or an accepted-answer set that excludes legitimate forms, its score distribution may differ for reasons unrelated to subject mastery.
H5P documentation describes basic score storage in certain host integrations and notes that detailed xAPI-based records depend on the hosting setup and capture mechanisms. Do not assume every embed stores learner results, or that every plugin exposes identical reports. The QA owner should verify the actual host’s result behavior with test attempts and inspect score, maximum score and completion state where available.
Use one correct attempt, one partly correct attempt and one incorrect attempt in each locale. Compare not only the number shown to the learner but also the gradebook or authorized reporting layer when the course relies on it. Translating labels must not create or erase evidence of completion, and assessment adaptations must be documented before results are treated as equivalent.
20. Separate interface translations from scoring configuration
H5P text overrides can change built-in phrases such as feedback headers or action labels. They should not be used as a workaround for incorrect scoring logic. A target “Correct!” message can look appropriate while appearing after the wrong selection because the correct-answer mapping was damaged. Conversely, an item may score properly but use an untranslated or misleading success label. These are separate failures with separate repairs.
Test scoring first using known answer choices, then assess whether the displayed feedback truthfully describes the result. For a partial-credit task, the language of feedback may need to distinguish “partly correct” from “fully correct.” Do not let an enthusiastic translation remove that distinction. A child needs precise evidence to learn from mistakes.
If scores are sent to an LMS, do not edit hidden integration identifiers in an attempt to translate them. Names intended for humans and keys intended for systems have different owners. Route any change to grading configuration through the platform specialist, then verify it using authorized test learner accounts.
21. Review copyright and reuse rights before distributing translations
Reusing an H5P package does not erase the copyright or licensing conditions attached to its media and authored content. A translated activity may include photographs, videos, music or diagrams governed by separate licences. Check that the intended use, alteration and redistribution are permitted, especially when moving content to another platform or publishing it openly. Record attribution where required and keep source information with the reused asset.
H5P’s Interactive Video tutorial demonstrates the use of a copyright information dialog for media. Treat those fields as part of the content record, not as expendable clutter. Translating an attribution statement does not authorize removal of an author’s name or license version. If a licensed source requires a particular attribution form, keep its identity and legal meaning intact while making any permitted explanation understandable to the target audience.
When a teacher creates original exercises around third-party video, separate rights to the quiz text from rights to the underlying media. The safest release is one that can explain where each asset came from and whether the translated derivative is allowed. A completed language review is not a substitute for permissions review.
22. Use AI-generated activity drafts with an explicit review boundary
H5P’s Smart Import documentation, updated in September 2026, describes generating activities in selected languages from source materials for eligible users. It also distinguishes fully supported language output from languages described as partially supported and warns that generated quality may vary. This capability can reduce initial authoring time, but it should not be treated as proof that a translated question preserves the original assessment construct or correct-answer logic.
If Smart Import is used, freeze the source learning objective and approved factual references first. Compare generated target questions with that objective, then review option validity, answer keys, explanations and scoring. A generated exercise is not necessarily a translation of an existing one; it may be a newly authored assessment with different examples. That difference matters when a school expects equivalent difficulty or comparable pre-test results.
The current Smart Import guidance also states that choosing a language applies to newly created imports and that existing Smart Imports are not simply switched afterward through that creation control. Check the installed product and plan before promising an automated language-change operation. Use the documented workflow that matches the actual environment.
23. Build a repeatable reviewer handoff
A translator needs more than a package file. Provide the activity purpose, age, subject vocabulary, source screenshots, intended answer mapping, scoring rule, approved alternatives and a list of content fields that are not to be edited. Where video is involved, provide a transcript and timing notes. Where an image has embedded text, provide the editable artwork or an approved replacement plan. This turns translation into an evidence-based task rather than reverse engineering.
The educator should receive the target version without the source answer visible, complete it as a learner, then inspect the answer key with the translator. This independent pass is especially valuable for negative questions, multiple correct answers and context-dependent vocabulary. Record whether any item required cultural or curricular adaptation. A small issue log with item ID, observed behavior, proposed repair and responsible owner is enough for many class-scale projects.
Have one technical reviewer test navigation and scoring in the intended host after the linguistic pass. The source and target may both work in a standalone authoring environment but differ inside an LMS frame, on a mobile device or under the actual reporting integration. Localization succeeds only when the intended learner can use the target activity where it will be assigned.
24. Worked example: a multiple-choice negative disappears
A Primary Science activity asks “Which material does not absorb water?” The source marks plastic as the correct option among several absorbent materials. In a translated version, the equivalent of “not” disappears while every answer choice remains correctly translated. A target-language learner correctly selects a sponge, but the activity marks the response wrong. The score now punishes knowledge instead of measuring it.
Repair the prompt to restore negation, then ask a native reviewer to solve it without seeing the source. Verify the configured correct flag and test both a correct and incorrect selection. Add the case to a small negative-scope regression set. This failure demonstrates why translation QA must inspect the instruction’s logical operators, not merely its topical vocabulary.
25. Worked example: correct options change after alphabetic sorting
An English geography exercise has four option records, the third marked correct. A translator sorts translated place names alphabetically by cutting and pasting text between editor rows, but the correct-answer flag remains attached to the original row. The target activity now rewards the wrong place. The source and target both display four plausible options, so casual screenshot review misses the error.
Restore the option-to-meaning mapping and reassign correctness only under educator review. Then shuffle runtime answer order if supported and complete the item twice. The important object is not the third visible line; it is the statement that should be correct. Position is presentation, while correctness is an instructional relation.
26. Worked example: a cloze answer rejects good grammar
A language exercise originally accepts the English base form of a verb. Its new Spanish sentence naturally requires a conjugated form, but the accepted-answer field still contains the translator’s dictionary lemma. Students enter the grammatically correct inflection and receive zero credit. The source author assumes pupils are careless because the score report shows many failures.
Rewrite the complete target sentence, define acceptable inflected answers and test them along with plausible wrong forms. Explain to the teacher which grammatical skill the cloze is now assessing. If the target language creates several legitimate completions, choose whether to accept them or redesign the prompt to remove ambiguity. A fair auto-scored exercise needs a target-language answer model.
27. Worked example: a video quiz appears before the translated caption
An Interactive Video presents a concept at 1:10 and opens a quiz at 1:15. The translated caption extends the sentence until 1:18 because the target needs more words. The quiz pauses playback before the pupil has finished reading the evidence. The target question may be accurate, but the activity now tests guessing or reading speed.
Adjust caption segmentation, interaction timing or pause behavior in accordance with the lesson’s design. Watch the target video from before the event and answer the question as a learner. Verify that changing timing does not hide later interactions or disrupt bookmarks. Translation length affects temporal structure when language is embedded in interactive media.
28. Worked example: a word puzzle becomes solvable by visual clues
An English vocabulary activity uses four options of roughly similar length. The target-language correct answer is a short common word, while three literal distractors become unusual phrases. Pupils can identify the intended answer without understanding the sentence. The activity appears successful because almost everyone scores full marks, but it no longer tests vocabulary meaning.
Have a target-language educator choose plausible alternatives that preserve the original misconception pattern. Keep the same learning objective while documenting the distractor adaptation. Pilot the item with a few learners and ask them to explain their reasoning. Unexpectedly high scores can be a clue to altered item difficulty, not simply excellent teaching.
29. Worked example: the water-cycle diagram remains English
A translated Science quiz uses Malay questions but its diagram contains English labels for evaporation and condensation. The pupil can answer if they know English, yet the activity was intended to assess Science through Malay. The visible caption below the image is translated, creating an inconsistent language experience without changing the answer key.
Replace or adapt the labelled artwork where permissions allow, preserve arrow direction and process identity, and check text alternatives. Then complete the item using only the target-language evidence. The test must prove that knowledge of the intended subject—not unannounced bilingual decoding—determines the answer.
30. Worked example: localized “Retry” creates a false promise
An H5P activity allows one scored attempt, but the target interface text override labels a navigation action as “Try again.” The button actually moves to the results screen rather than allowing another scored attempt. The mistranslation changes a learner’s expectation about assessment rules even though score calculation is correct.
Inspect the real action behind the control and restore a label that describes it truthfully. Test at both successful and unsuccessful outcomes, since interfaces may show different controls by state. Translators should not choose button wording without knowing what happens when the learner activates it.
31. Worked example: scores exist in the activity but not in the LMS
A teacher translates a quiz, exports it as an H5P package and embeds it on a course page. Students can complete it and see scores, but the instructor’s gradebook remains empty. The teacher assumes translation damaged scoring. The actual issue may be the host’s result-capture mechanism: a standalone embed is not equivalent to an assessed LMS activity with an integration designed to store grades.
Verify the hosting method and authorized reporting configuration before editing question text. Complete a safe test attempt in the target environment and inspect whether score and completion reach the correct course record. H5P documentation distinguishes generated interaction statements from the platform mechanisms that capture and store them. Translation should preserve assessment meaning, but hosting must provide the evidence pipeline.
32. Worked example: two translated variants assess different curricula
A Mathematics activity about ratio is translated for two regions. One reviewer converts the word problem from litres to gallons and rounds every quantity; another retains the original metric values. The resulting numerical relationships are no longer identical, so score comparisons across classes become misleading. Both activities are grammatically excellent but not equivalent assessments.
Either preserve the numbers and units, or redesign the adapted problem with mathematically exact relationships and educator approval. Record whether equivalence was maintained and what form it takes. The stronger principle is transparency: curriculum adaptation can be valuable, but it must not be disguised as a literal translation when reporting results together.
33. Preflight checklist for a translated H5P activity
Check that the source is approved and preserved; the copy retains content-type behavior; the target objective is written explicitly; all learner-facing fields are translated or deliberately exempted; question instructions preserve qualifiers; correct-answer mappings match translated meaning; distractors remain plausible; cloze answers accept intended forms; feedback teaches the right rule; interactive timing is usable; embedded media and alternatives are localized; and the activity functions under the target host.
Then run answer-path tests. Intentionally answer one item correctly and incorrectly, complete a multi-answer or partial-credit case when relevant, and inspect feedback, score and completion. For video or drag-and-drop, test on an actual device with the expected learner interactions. For assessed courses, verify the host’s authorized results workflow rather than relying solely on the number displayed in the activity.
Finally, check attribution and distribution rights, source and target revision labels, curriculum scope, reviewer sign-off and a destination link that opens the right activity. One last learner walkthrough often finds the most important defect: a sentence that is technically correct but instructs an impossible action in the localized interface.
34. Transfer exercise: rebuild a Primary Science item for a new language
Imagine an English H5P Multiple Choice item: “When water vapour cools, which change of state occurs?” Four options are displayed; condensation is correct. The target is a bilingual Primary Science class taught primarily in Malay. Before translating, write the intended concept and misconception: cooling water vapour produces liquid water, not evaporation. Provide a diagram with no embedded English labels or a properly localized one, and choose terms aligned with the teaching materials.
Draft the Malay prompt and options, then ask a target-language Science teacher to solve the question independently. Check that the options are parallel in grammatical form, the correct-answer flag remains attached to condensation, and wrong-answer feedback explains cooling rather than simply revealing the right word. Run the exercise on a tablet with large text. Verify both score and the learner’s ability to understand why an answer is right.
Now change one assumption: the activity becomes an Interactive Video with captions and the question appears just after a water droplet forms. The same conceptual translation must also survive timing and media synchronization. This transfer shows why the answer key is only one piece of assessment validity. The content, interaction and evidence must agree.
35. What makes this workflow genuinely faster for tutors
A tutor loses time when every new language requires rediscovering which fields are translatable, which answers are correct and which settings control results. A reusable item manifest, source revision, option mapping and answer-path test prevent that rework. Instead of translating quickly and correcting mistakes in front of pupils, the educator can validate one small representative activity and scale a reliable method to the remaining lesson.
The deepest improvement is a distinction between language and learning mechanics. Translators need freedom to form natural sentences, substitute an equivalent cultural example and choose appropriate grammar. They should not alter scoring flags by accident. Developers and platform administrators protect package compatibility and reporting. Educators decide whether the target task still measures the intended skill. When those responsibilities are clear, review becomes quicker and more meaningful.
Summary: translation is finished when the learner can succeed fairly
A localized H5P activity succeeds when the learner receives understandable prompts and feedback, can respond using the intended language, is scored according to valid target-language answers, and can see results that reflect the assessed learning objective. Changing built-in buttons is useful, but it does not translate a quiz; copying a package is useful, but it does not prove answer equivalence; and generating a target-language activity is useful, but it does not remove educator review.
Keep this URL focused on H5P assessment integrity across languages. Parent guides to language learning, platform-wide Moodle localization and broader authoring-tool QA should link here when they need this specific lesson, not compete with it by recreating another generic H5P overview.
Frequently asked questions
Can I translate an H5P activity automatically?
Tools may help generate or adapt text, but correct answers, hints, branching, accepted input and learning objectives still require a target-language educator’s review. Automation of text is not automatic validation of assessment meaning.
Does the H5P language switcher translate my quiz questions?
No. It concerns built-in interface text available through supported authoring options. Author-written prompts, options and feedback must be reviewed and translated separately.
Can I download and reuse an existing H5P file?
Where reuse is enabled and permitted, H5P documents a workflow for downloading a .h5p file and uploading it into a compatible authoring environment. Permissions, host features and library compatibility must be checked.
How do I keep multiple-choice scores correct after translation?
Map each source option to its target meaning and preserve or deliberately reassign correct-answer settings. Complete both correct and incorrect test attempts, including shuffled options when applicable.
Why does Fill in the Blanks mark a valid translated answer wrong?
The target sentence may require different inflection, spelling or legitimate alternatives from the source. Revise the accepted-answer settings after writing the natural target-language sentence, and test realistic inputs.
Will an embedded H5P quiz automatically send grades to my LMS?
Not necessarily. Result storage depends on how the activity is hosted and integrated. Verify actual gradebook or authorized reporting behavior in the target environment.
References and further reading
- H5P: Language switcher — built-in interface text and authoring language options.
- H5P: Download or copy content using Reuse — documented reuse and package handoff workflow.
- H5P: Interactive Video tutorial — interaction timing, media, quizzes and authoring guidance.
- H5P: Analyzing results and answers — host-dependent results and reporting considerations.
- H5P: Smart Import language support — 25 September 2026 guidance on generation-language features and limits.
- eduKateSG: How People Translate Quickly series index — neighboring platform-specific translation QA articles.