eLearning Translation: What It Costs, How Long It Takes, and How to Get It Right

Nobody can tell you what it really costs to translate a 60-minute course into five languages, which makes it very difficult to ask for the budget. 

In this guide, we’ll show you how to find per-word rate ranges by language tier, voiceover and subtitling costs per finished minute, engineering fees associated, and a fully itemized example for a 60-minute Storyline course into five languages. 

Plus any realistic timelines, framework for when AI translation is really safe to use, and what actually breaks when you export from Storyline, Rise, or Captivate. 

Let’s get started. 

eLearning Translation vs. Localization vs. Transcreation

These three terms are usually used interchangeably, but they’re cost differently. 

What does eLearning translation actually cost? 

Everything listed below is a market range, so rates will vary by vendor tier, volume, and how clean your source files are. Translation sits on top of your build budget, so it helps to know your baseline eLearning development costs before you layer language counts on top. You can use these ranges to sanity check quotes you’ll receive and to size a budget request, then validate against 2 or 3 real bids from top eLearning translation companies.

Per-word translation rates by language tier

Assumes professional human translation with a second-linguist review, no translation memory discount on the first project. Rates here will drop 15–30% on repeat work once a TM is built, which is why TM ownership matters.

You can count your words before you ask for a quote – like a 60-minute course typically runs 6,000 to 9,000 words once you combine on-screen text and narration script.

Voiceover and subtitling

Priced per finished minute of course runtime, per language.

Human VO is almost always the single largest line item in a localization budget, so it is also the first place to cut – subtitling a course instead of just dubbing it will typically remove 60 to 75% of the total cost. 

Engineering and Rebuild Fees

Someone has to actually pull the translated strings back into the course, fix layouts that broke under text expansion, re-sync captions, re-record timings, and republish the SCORM package. 

• Hourly: $60 to $120 per hour
• Per slide: $8 to $20 per slide (a 60-minute course is usually around 60 to 90 slides)
• RTL languages (Arabic, Hebrew): add 30 to 50% – as layout inversion is manual work
• In-country linguistic review: $400 to $900 per language
• Project management: 10 to 15% of project total

Example: 60-minute Storyline course into five languages

To help you see exactly how much an eLearning costs (approximately), here’s a sample.

Assumptions here are: 

• 7,500 source words
• 75 slides
• Target languages: Spanish, French, German, Japanese, Arabic.

 

It’s worth remembering that the budget roughly is $80 to $110 per finished minute, per language for a subtitled course, and $300 to $400 per finished minute, per language for full human voiceover. Multiply that by runtime and language count for a defensible first-pass estimate. 

How Long eLearning Translation Takes?

Timelines below assume our working example: a 60-minute course, clean source files, and a vendor with capacity. You can add 1 to 2 weeks if you’re onboarding a new vendor or building a glossary from scratch, the same ramp-up you’d budget for any eLearning outsourcing engagement.

These are five languages in parallel that will cost you roughly 50% more calendar time than just one – not 5X. So run them together, or you’ll add months (more expensive cost on your end). 

Two Things to Consider That Actually Blow the Deadline

• In-country SME review – in practice, this is actually unpaid work just sitting on top of someone’s day job in another timezone, and it may routinely stretch to 3 to 4 weeks. So you really have to name your reviewers before any kickoff, just simply book a time in their calendars and give them a specific deadline.

1. Voiceover re-records – any script change after every recording means you’ll return to the studio, and most studios book out. Lock scripts hard at the translation approval gate. For instance, a single late terminology change can change across 5 languages, which will cost you 2 weeks. 

 

AI, Machine Translation, or Human?

Here’s the thing: machine translation is genuinely good enough for a meaningful share of corporate training content, and conversely genuinely dangerous for the rest. So the question isn’t which is better – it’s which tier each course earns. It’s the same judgment call teams are already making with AI for instructional design: the tool is fine until the stakes of getting it wrong outrun the savings.

Three Quality Tiers

MTPE is where most of the market is landing here, and it’s the tier vendors are really quietly using whether or not they say so. 

Decision Matrix By Content Type

Where AI Genuinely Breaks

As we know, AI isn’t perfect at all. So you really have to put guardrails and review checkpoints when you’re using AI in eLearning translation. Here are a couple of elements to check:

• Assessment feedback strings – “not quite” – “try again” and “incorrect” can carry different weight in different cultures, and MT flattens that distinction.
• Idiomatic scenarios – let’s say, a workplace conflict scenario built on American directness norms doesn’t translate – it needs actual rewriting. Branching decision paths and dialogue in scenario-based eLearning services are the most expensive content to localize for exactly this reason.
• Legal and policy phrasing – terms of art that have jurisdiction-specific meanings that MT can confidently get wrong.
• Glossary drift – across a 20-module library, for example, MT will translate the same product term 4 different ways unless a TM and termbase are truly enforced. 

Tool-Specific Workflows and What Breaks

This isn’t a comprehensive process on how to use AI tools for eLearning translation – as each tool demands another how-to guide on its own. How painful the export-translate-reimport loop gets depends heavily on which of the best eLearning authoring tools your library was built in. But to give you quickly how each works, here we go:

1. Articulate Storyline

You can export via Translation → Export to XLIFF 1.2/2.0 (better to help preserve segment structure and works really well with any TM tool), or you can Export to Word (which is easier for reviewers, but worse for tooling). 

Then translate, then reimport into a duplicated copy of the project. 

What it really doesn’t export:

• Text baked into imported images or SVGs
• Text inside embedded video
• Alt text on some object types, depending on version
• Custom JavaScript strings
• Text typed into the player’s custom labels unless exported separately (Player → Text Labels)

The recurring challenge here is fixed-width containers. Storyline, for instance, has text boxes that don’t reflow, so German and Russian expansion overflows the shape or clips at the boundary. If most of your library is Articulate eLearning, assume a slide-by-slide visual pass is part of every language, not an exception.

Autofit essentially can help and also silently shrink type below legible size – check every slide visually after reimport. 

2. Articulate Rise

Rise can export XLIFF, and reimport is block-level. So it’s cleaner than Storyline given that Rise is responsive – text expansion mostly handles itself. 

Where AI can break is only the knowledge check feedback; labeled-graphic markers, sorting activity labels, and button text can sometimes come back incomplete. Any block that’s edited after export can lose its mapping on reimport. You can freeze the source course at export and simply don’t touch it. 

3. Adobe Captivate

You can export via File → Export → Project Captions and Closed Captions for captions, plus XLIFF for slide content. So for two things it will reliably break: one is variables ($$name$$ placeholders get translated or mangled, breaking the substitution) and interaction states – hover, visited, and disabled label text can often live outside the export. 

Captivate can also help build sentences from concatenated variables, which grammatically collapses in inflected languages. So flag all these to the translator explicitly with a do-not-translate list. 

4. Camtasia and Video Assets

You can export captions as SRT or VTT, translate, reimport. Pretty simple and straightforward unless there’s any burned-in text or on-screen callouts, in which case you’re just re-rendering the video per language, not just translating it. So that’s a production cost issue here. 

Translated captions can run longer than English, so reading speed here may drop below comfortable. 

Here are also common breakage points you can check every one after reimport: 

• Text expansion overflowing fixed-width containers (German, Russian: +30 to 50%)
• Caption and voiceover desync after script length changes
• Quiz feedback and results-slide strings that are reverting to English
• Text baked into images and screenshots
• RTL layout inversion for Arabic and Hebrew-like navigation, progress bars, and slide order all invert
• Fonts without full character support (CJK, Cyrillic, Arabic) that are rendering as boxes
• SCORM completion and tracking breaking on republish

Prep your source files before you translate anything

Most localization overruns aren’t essentially translation problems – they’re just source-file problems that were cheap to fix in English and expensive to fix 5X over.

Write for translation

• You can keep sentences under 25 words and one idea each
• You can cut idioms, sports metaphors, and humor that only depend on wordplay
• Simply avoid culture-bound examples likenames, holidays, legal references, currency, imperial units – if not needed.
• You can use consistent terminology; don’t vary phrasing for style (“learner,” “user,” and “employee” become three different words in five languages.
• You can write dates as unambiguous formats, and never as numerals alone

Externalize every string

• No text baked into images, screenshots, or diagrams – simply use a text layer over the image.
• You can keep on-screen text in native text objects, not embedded video
• You can pull player labels, button text, and quiz feedback into the export before you assume they’re included
• Simply keep editable source files for every graphic – if you can’t edit the PSD, you’ll just rebuild it.

Design for expansion

• Just leave 30 to 50% empty space in every text container
• Simply avoid fixed-width buttons sized exactly to English labels
• Start testing the worst case early: drop German placeholder text into your three most crowded slides before you commit to the template
• Simply choose fonts with CJK, Cyrillic, and Arabic coverage

Lock and document before kickoff

• Start building a glossary of 30–80 key terms with definitions and do-not-translate items (product names, UI strings, legal terms)
• Start writing a one-page style guide: tone, formality register (critical for languages with formal/informal address), abbreviation rules.
• Start the source course – any English edit after extraction multiplies across every language.

Post-Translation QA

Every eLearning solution requires quality assurance, and you can run these 3 separate passes for post-translation QA. 

1. Linguistic review (in-country)

This is where a native speaker in the target market reads the course as a learner would: in the published build. So checking terminology consistency against a glossary, formality register, cultural fit of scenarios, and imagery, and anything that will read as translated rather than just written. You can budget 2 to 4 hours per language for a 60-minute course. 

2. Functional testing (LMS)

You may republish the SCORM package and test it in your actual LMS (not in your authoring tool preview). If tracking breaks on republish, it’s usually a packaging issue rather than a translation one – worth reviewing how to create SCORM content that survives repackaging across language variants. Simply check for completion and pass or fail tracking fires correctly, bookmarking resumes, quiz scoring reports, links resolve, and a course that will launch in browsers – your learners actually can use.

3. Media QA

This is where caption timing and reading speed, voiceover sync against animations and slide advances – no clipped or overflowing text, fonts that are rendering all characters, RTL layouts inverted correctly throughout. Caption accuracy, legible type size, and correct reading order are also the baseline for accessible eLearning, so run this pass against your accessibility checklist in the same sitting.

Final Thoughts

eLearning translation isn’t just expensive with the words – it’s more costly given that everything around them: voiceover, rebuild, and review cycles. Get your source files clean, match each course to the quality tier it actually needs, and start budgeting the engineering line before you shortlist from the top eLearning vendors.