Answer
eLearning Localization: How Translators and Agencies Produce Multilingual Courses
eLearning localization is how translators adapt courses, SCORM packages, Storyline files, voiceover, and LMS content for each target locale from the production
eLearning localization is a production lane translators and agencies run alongside How App Localization Works and What Clients Are Buying, converting a source course into release-ready builds per locale. The lane handles slide text, narration, graphics, interactions, and SCORM or xAPI packaging through CAT tools, translation memory, term bases, and in-LMS review. It sits inside the broader program described Localization Management: How Translators and Agencies Run Multi-Locale Programs.
Key points
What is eLearning localization from the translator's side of the desk?
eLearning localization is the end-to-end adaptation of an online course — slide text, narration, graphics, interactions, and SCORM or xAPI packaging — into a release-ready build per locale, produced by translators and agencies using CAT tools, translation memory, term bases, and in-LMS review before handover. The lane sits between authoring and the client's LMS, and its deliverable is a package the LMS can import without manifest errors or broken scoring.
Buyers see a localized course. The translator sees seven artifacts, an XLIFF round trip, a voiceover branch, a layout risk, and an LQA scorecard. That is the production reality this page documents.
What is the terminology map for translation, localization, and adjacent lanes?
| Term | What it is | When you need it |
|---|---|---|
| eLearning translation | Converting slide text and transcripts into the target language | Short internal modules where layout and narration stay in English |
| eLearning localization | Full adaptation of text, audio, graphics, interactions, and SCORM metadata per locale | Any course released to learners in the target market and tracked in an LMS |
| Certified translation | Attested translation with a certificate of accuracy per page | Regulated training audited by USCIS-adjacent, legal, or medical reviewers |
| Multimedia localization | Voiceover, subtitling, and on-screen graphic re-render | Narration-driven courses or marketing-style modules |
How does eLearning localization differ from plain translation?
Translation converts source text; eLearning localization also adapts narration audio, on-screen graphics, cultural references, assessment items, SCORM metadata, and slide layouts so the course runs correctly inside the learner's LMS. Plain translation ends at the word. Localization ends at the LMS gradebook posting a quiz score in the target locale.
- Narration is re-recorded or re-timed, not re-typed.
- Graphics with embedded text are re-rendered rather than string-replaced.
- Quiz distractors are re-numbered when examples change, and answer-key IDs stay locked.
- SCORM manifest metadata is updated per locale without touching tracking IDs.
- Slide layouts are re-fitted around text expansion before QA.
Where does eLearning sit inside the agency's localization program?
eLearning is one lane inside a multi-lane program alongside Product Localization: How Translators and Agencies Deliver It, Website Localization: What Translators and Agencies Actually Do on the Project, content, legal and compliance, and in-product video. The lanes share the term base and style guide. eLearning adds SCORM or xAPI packaging, narration recording, and LMS QA steps the other lanes do not run.
Certified translation is a seventh regulated lane sitting alongside the six. On regulated training it layers onto the eLearning deliverable, carrying its own competent-translator certificate requirement.
Why does cultural and pedagogical adaptation go beyond swapping words in a slide?
Effective localization rewrites examples, humor, and scenario characters to match the learner's cultural frame, and adjusts instructional sequencing where local learning norms differ. A technically accurate course that keeps US-only names, holidays, and workplace scenarios still fails knowledge-transfer goals in Germany, Japan, or Saudi Arabia.
- Scenario characters renamed and re-cast per locale.
- Workplace examples swapped for ones the learner recognizes.
- Humor and idiom replaced where a direct render reads flat.
- Sequencing re-ordered where local pedagogy expects context before task.
What are the key components of an eLearning course a translator has to localize?
A translator localizes seven components: slide text and quizzes, narration audio, on-screen graphics, interactions and simulations, SCORM or xAPI metadata, LMS interface strings, and locale-specific legal and compliance notices. Each component has its own tool chain, its own QA pass, and its own place in the delivery timeline. Because these components are interdependent — a mistranslated quiz stem breaks both the learner experience and the LMS gradebook — the translator must treat the course as a system, not a document.
flowchart TD
A[Source course in Storyline or Rise] --> B[XLIFF export]
B --> C[CAT translation against TM and term base]
C --> D{Narration branch}
D -- Voiceover --> E[Studio recording and re-timing]
D -- Subtitle --> F[SRT or VTT per locale]
C --> G[Graphic re-render]
E --> H[SCORM re-packaging]
F --> H
G --> H
H --> I[LQA against MQM scorecard]
I --> J[In-LMS review]
J --> K[Release to client LMS]
Which source files land in the translator's workspace at kickoff?
The translator receives seven production artifacts at kickoff: the Storyline or Rise source file, an XLIFF export, the SCORM or xAPI package, the narration script with timecodes, a subtitle file in SRT or VTT, the locale style guide, and the client term base. Each artifact is mandatory — the XLIFF carries slide text and quiz items into the CAT tool, the narration script maps spoken lines to slides for re-timing, and the term base locks product names and regulated terms so they survive every translation pass unchanged. Arriving at kickoff without any one of these artifacts forces a scope-change before translation starts.
| Artifact | Format | What it carries |
|---|---|---|
| Authoring source | .story,.rise,.cptx | Editable slides, interactions, timeline cues |
| XLIFF export | .xlf | Slide text and quiz items for the CAT tool |
| SCORM or xAPI package | .zip | Manifest, tracking IDs, launch files |
| Narration script | .docx with timecodes | Spoken lines mapped to slides |
| Subtitle file | .srt or.vtt | Timed captions for the narration track |
| Locale style guide | .pdf or.docx | Tone, register, formatting per locale |
| Term base | .tbx or CSV | Product names, regulated terms, locked translations |
What do multimedia assets add to the translator's queue?
Multimedia assets add narration audio, timed subtitle files, screen-recorded demos, animated graphics with embedded text, and interaction scripts, each routed through the same translation memory and term base but recorded, re-timed, or re-rendered per locale before SCORM re-packaging. Animated graphics are particularly labour-intensive because embedded text is not extractable through XLIFF — it must be listed in the file audit, sent to a graphic designer per locale, and re-rendered before it enters the package. Voiceover localization is the highest-cost multimedia element: full voiceover with studio direction costs several times what subtitle-only production costs per source minute, and the branch chosen reshapes the entire delivery timeline. The subtitle-only path runs approximately $7–
How does the core eLearning localization workflow run from kickoff to LMS publish?
1
What does the file audit catch before translation starts?
The audit catches locked slide containers, hard-coded strings outside XLIFF, embedded text on graphics, un-timed narration cues, missing subtitle files, SCORM manifest errors, and placeholder patterns that break in the CAT tool. Each finding is logged and resolved before the XLIFF is handed to translators, because a hard-coded string discovered mid-translation forces a scope change that delays the entire locale. The audit also confirms that the SCORM manifest declares the correct LMS target version, since a version mismatch breaks tracking on launch regardless of how accurately the text was translated.
- Fixed-width text boxes flagged for auto-fit or redesign.
- Hard-coded strings extracted into XLIFF.
- Embedded graphic text listed for re-render per locale.
- Narration timecodes verified against the subtitle file.
- SCORM manifest validated against the LMS target version.
- Placeholder patterns (variables, learner names) locked as non-translatable.
2
How does text expansion affect Storyline and Rise slide layouts?
Text expansion reshapes layouts because German, Finnish, and Russian expand slide text 20–35% longer than English, breaking fixed text boxes and button widths unless the translator flags those containers at XLIFF import. Articulate Storyline is the primary authoring tool involved in this step: its fixed slide containers do not auto-resize by default, so either the master slide must be redesigned to accommodate the longest target locale or auto-fit must be enabled on every affected text box before re-packaging. French adds 15–20% and stretches callout bubbles; Chinese Simplified, by contrast, runs 20–30% shorter than English and requires re-tuning vertical rhythm so containers do not look sparse. Identifying all affected containers during the file audit — before translation starts — is the only reliable way to prevent layout rework after XLIFF import.
| Locale | Expansion vs English | Impact on Storyline containers |
|---|---|---|
| German | +20–30% | Button labels overflow; headers wrap |
| Finnish | +25–35% | Compound words break fixed boxes |
| Russian | +20–30% | Headers overflow; Cyrillic kerning widens lines |
| French | +15–20% | Longer phrasing stretches callouts |
| Chinese Simplified | -20 to -30% | Containers look sparse; vertical rhythm needs re-tuning |
3
When does voiceover run versus subtitle-only?
Voiceover runs when the course is narration-driven, compliance-mandated, or targeted at audio-first learners; subtitle-only runs when budgets are tight, update cadence is high, or the audience accepts captions. Voiceover costs several times subtitle-only per source minute and reshapes the delivery timeline per locale because studio booking, direction, recording, and re-timing are sequential. A course with a high update frequency is a strong candidate for subtitle-only because new subtitle files are produced and swapped in days, while a new voiceover recording requires studio availability and re-timing before re-packaging. The compliance or regulatory context is the deciding variable for narration-driven training: when an auditor or regulator expects the learner to hear the target language, subtitle-only is not sufficient.
What tools and authoring platforms do translators use for eLearning localization?
Translators use four tool categories: authoring tools (Articulate Storyline, Rise 360, Adobe Captivate), CAT and TMS platforms (Phrase, memoQ, Trados) handling XLIFF, SCORM editors for re-packaging, and audio production tools for narration — all wired to the same translation memory and term base.
| Category | Tools | Role in the lane |
|---|---|---|
| Authoring | Articulate Storyline, Rise 360, Adobe Captivate | Produces source and regenerates the course after translation |
| CAT and TMS | Phrase, memoQ, Trados | XLIFF handling, TM and term base enforcement |
| SCORM editor | Reload Editor, SCORM Cloud toolset | Manifest fixes and package validation |
| Audio production | Studio DAW, subtitle timing tools | Narration recording, SRT/VTT timing |
Where the LMS front-end is built in React, UI strings route through an i18n pipeline documented React i18n: How Translators Fit Into a React Localization Pipeline. Where course content lives inside a WordPress or headless stack, the content side is handled per CMS Localization: How Translators and Agencies Produce Multilingual Content Across Platforms.
How do Storyline and Rise exports flow through the CAT tool?
Storyline exports an XLIFF of all slide text and quiz items that the translator loads into Phrase, memoQ, or Trados, works against the TM and term base, and re-imports to regenerate the.story or Rise file with layouts intact when text-expansion flags were respected.
Where do machine translation and AI fit in the eLearning lane?
Machine translation and AI drafts fit as a pre-translate layer inside the CAT tool for high-volume slide text and transcripts, always followed by human post-editing and the LQA gate. Raw MT output is never shipped into a course carrying an LMS gradebook or a compliance record.
How do translators protect SCORM and xAPI integrity through localization?
Translators protect SCORM and xAPI integrity by keeping tracking IDs, bookmark state, quiz scoring, and manifest metadata untouched inside XLIFF, re-packaging through the authoring tool rather than editing the ZIP directly, and running an LMS smoke test on every locale. SCORM integrity is not a post-production check — it is a constraint enforced at the CAT tool level, at the re-import step, and at the smoke-test step, in that order. Any one of those three enforcement points can catch a different class of breakage: the CAT tool catches accidentally translated identifiers, the re-import catches manifest mismatches, and the smoke test catches runtime tracking failures that only appear when a real LMS processes the package.
Direct ZIP edits are the most common source of tracking breakage in localized eLearning. Editing a SCORM ZIP outside the authoring tool — to force a layout fix or update a string — can corrupt the manifest, shift interaction identifiers, or orphan bookmark state entries that the LMS relies on to resume a learner's session. The correct path is always to re-import the translated XLIFF through Storyline, Rise, or the equivalent authoring tool, regenerate the package, and then update manifest language metadata to reflect the target locale before exporting the final ZIP.
xAPI packages carry the same risk as SCORM in a different format. Statement actors, verb IDs, and activity IDs are all defined in the package and must remain stable across the localization round trip. The translator locks these fields as non-translatable in the CAT tool at the same time as SCORM tracking IDs, and the re-packaging step regenerates them from the authoring tool's source rather than from the intermediate XLIFF state.
- Lock tracking IDs and interaction identifiers as non-translatable in the CAT tool.
- Translate only the fields the authoring tool externalized to XLIFF.
- Re-import through Storyline, Rise, or Captivate to regenerate the package.
- Update manifest language metadata to the target locale.
- Smoke-test launch, resume, quiz scoring, and completion in a sandbox LMS.
How is quality assured on a localized eLearning course?
Quality is assured through a two-stage gate: linguistic review inside the CAT tool against the term base and style guide, then an LQA sample at $40–$90 per 1,000 words scored on an MQM-style scorecard aligned to ISO 17100 covering accuracy, fluency, terminology, locale convention, and style.
What does the LQA scorecard check on a localized course?
The LQA scorecard checks five MQM defect categories — accuracy, fluency, terminology, locale convention, and style — plus eLearning-specific checks for narration-to-text sync, subtitle timing, and quiz-item parity against the source.
Why can't compliance courses skip the LQA gate?
Compliance courses cannot skip the LQA gate because regulated training is audited against frameworks like ISO 17100, and skipping it is a compliance gap the client's auditor flags at the next review. The gate is not a cost line to cut; it is the evidence trail.
Pricing
How do agencies price eLearning localization?
Agencies price eLearning localization across four primary line items: per-word translation of slide text and transcripts, per-minute voiceover or subtitling of narration, per-hour engineering for SCORM re-packaging and layout fixes, and the LQA gate at $40–$90 per 1,000 words on regulated lanes. These line items are additive — a single locale delivery carries all four — and the total project cost is therefore a function of word volume, audio minutes, engineering complexity, and whether the course is regulated. Regulated training adds a fifth line item: the certified translation lane, which prices separately from standard localization and requires a certificate of accuracy on every page.
How we run the eLearning lane
We produce first drafts of certified translations and feed them into the eLearning lane as the attestation layer on regulated training. On a localized course, we draft the certificate of accuracy per page, lock the regulated terminology in the term base, and route the output through human review and LQA before any file leaves the lane. The LQA scorecard we apply covers accuracy, fluency, terminology, locale convention, and style — the five standard MQM categories — plus the three eLearning-specific categories that matter to an auditor: narration-to-text sync, subtitle timing, and quiz-item parity between the source and target version of every assessed interaction.
We decline to raw-ship machine-translated output on a compliance course. The LQA gate exists because an auditor will read the gradebook and the attestation together, and raw MT fails that read on terminology and locale convention consistently enough that passing the gate requires human post-editing at a level that closes the cost gap with full human translation. We price LQA in; we do not offer a lane that omits it on regulated content.
We decline to edit SCORM or xAPI ZIPs directly to force a layout fix. Direct ZIP edits break tracking IDs, manifest metadata, and bookmark state in production more often than they save time — the breakage simply moves from the engineering step to the client's LMS environment, where diagnosing it costs more. We re-import through the authoring tool, regenerate the package, update the manifest language metadata to the target locale, and run the smoke test against a sandbox LMS on every locale delivery, even when the schedule is tight. Text-expansion issues in German, Finnish, Russian, and other high-expansion locales are addressed at the authoring level — auto-fit or a resized master slide — not by editing the exported ZIP.
We do not claim universal LMS acceptance. SCORM and xAPI packages behave differently across LMS versions, and a package that launches cleanly on one platform fails on another due to manifest version declarations or JavaScript runtime differences. We test against the specific LMS version the client runs, and we direct clients to confirm with their platform administrator before publish rather than promising a build will clear every system.
Translators and agencies ask five recurring questions: tool choice, formatting after translation, localization-versus-translation scope, RTL handling, and the specific challenges of Storyline and Rise courses.
Related
- i18n Translation: How Translators Produce Localized Builds from JSON, YAML, and XLIFF
- i18n-editor: What Translators Use to Edit JSON Translation Files
- i18n: What Internationalization Means for Translators Producing Localized Builds
- memoQ as a CAT Tool: Features, Licensing, and Certified-Lane Fit
- Guides for translators and translation agencies
- Certificate of translation accuracy: what it must say