When a training video needs another language, the first question is often whether to translate the subtitles or replace the narration. The better starting point is your learning goal or what learners must understand while the video plays. Based on your goal, what is the communication method? Are they following a demonstration, reading a busy screen, listening on headphones, or watching in a shared classroom?
The answer shapes the language work: script translation, spoken delivery, timed text, and changes to the video itself. A translated script can be accurate, yet too long for the scene. A clear voice can still incorrectly pronounce a product name. Subtitles may localize the spoken content, while leaving important screen text untranslated. These decisions should be made before production begins and shared with your translation provider in a project specification brief.
Start with the Audience and Viewing Conditions
Define the target languages and regional variants, learners’ familiarity with the subject, and the setting in which they will watch. Do not infer reading ability, hearing needs, or preferred delivery from a language or nationality. Also ask the trainer about their audience insight, including people who need alternatives to the audio or visual information.
Identify whether learners can pause and replay independently as needed. Someone following a self-paced lesson has different viewing conditions from a group watching a fixed presentation. Where the video shows a detailed action, learners may need time to examine the image as well as read translated text. This is an important consideration to discuss when creating the localized version.
Also ask what will remain in the original language. A software demonstration may show English menu names that users must recognize, while the explanation is translated. Decide whether the interface will be replaced, described using its existing labels, or supported by a terminology reference. Narration and subtitles alone do not translate every visible element.
Compare Three Practical Delivery Options
Original Audio with Subtitles
This approach retains the original speaker and provides written translated subtitles. The most logical reason to keep is when there is a limited budget but timing available to read captions, or when preserving the presenter’s original delivery takes precedence. However, keeping the soundtrack does not remove the translation, cue timing, and review considerations. Dense explanations or rapid scene changes may also require adaptation.
Check whether the localized text competes with the demonstration, diagrams, or existing screen labels. If the original video has subtitles permanently embedded in the picture, a second language can create confusion with overlapping text. Request a clean video export where possible. If none exists, discuss the treatment of the original text before deciding on this option.
Translated Narration
A new audio track lets learners hear the explanation in the target language. It may suit material that needs sustained attention to the visual demonstration. It also introduces voice selection, pronunciation checks, audio review, and synchronization. Replacing narration does not remove the need to consider captions or other accessibility provisions.
Clarify whether the original speech should disappear or remain quietly underneath. If speech, music, and relevant sounds exist only in one mixed track, replacing the voice may involve additional editing or a revised approach. Ask for separate audio tracks when available. Do not assume that a finished video contains all the source material needed for a clean replacement.
Translated Narration with Matching Text
Providing both can support different viewing needs, but the two outputs must be coordinated. Text that accompanies the new voice should follow the final target-language audio, rather than an earlier script draft. Agree whether the text is a subtitle translation, accessibility captions, or both in separate tracks. Define the deliverables by function, not only by the word “subtitles.”
Record the decision for each audience in a simple delivery list: language variant, retained or replaced audio, required text track, treatment of screen text, and playback destination. One course can have different justified arrangements for different audiences. The aim is a clear production scope, not uniformity for its own sake. Have the training owner approve that list before translating the entire library.
Distinguish Subtitles, Captions, and Transcripts
Terminology varies across regions. W3C distinguishes translated subtitles from captions that convey speech and relevant non-speech audio, such as an essential sound or speaker identification. A translation of dialogue alone may therefore be insufficient for someone who cannot hear the soundtrack. Specify the information required in each text track.
A transcript is a separate text version, useful for reference and review. A descriptive transcript can also communicate important visual information. It should not be treated as an automatic replacement for synchronized captions. Ask the training or accessibility owner to define the applicable requirements for the course and its delivery setting.
W3C guidance on captions and subtitles
Likewise, translating spoken narration does not necessarily make the visuals understandable to someone who cannot see them. If the instructor says “select this” while pointing, necessary information may exist only in the picture. W3C describes ways to integrate visual description into narration or provide it separately. Any added explanation needs approval for accuracy and timing.
W3C guidance on describing visual information
Prepare a Script That Can Be Spoken and Timed
Provide the approved video and its matching script, with scene or segment identifiers. Mark narration, on-screen text, and production notes separately. If no script exists, agree how the spoken source will be transcribed and checked. An old script that no longer matches the edit can create more confusion than a clearly identified transcription task.
Translate for the intended use. Spoken explanations need natural phrasing, while timed text needs readable segmentation. Preserve the instruction, qualifications, and terminology when adapting either. If a sentence cannot fit the scene without losing meaning, raise the issue with the content owner. An editor should not remove a condition or warning simply to match the original duration.
Record visual cues beside the relevant script segments: a button appearing, a diagram changing, or a pause for an exercise. This lets the team judge timing against the learning sequence. For screen demonstrations, use approved interface names and confirm whether the target-language interface differs from the recorded version.
Identify the approved video edit by filename and revision, and freeze it for the agreed production stage. If editing continues, provide a change log rather than quietly replacing the reference file. Even a short inserted introduction can shift all subsequent text cues. Agree who notifies the language team and how the impact of each change will be assessed.
For example, consider a fictional onboarding lesson showing how to submit an expense claim. The translated explanation takes longer than the original screen action. The owner might approve a longer hold on the relevant screen or a more concise explanation that preserves the required steps. Speeding up the voice is not the only production choice.
Evaluate AI Narration with Your Actual Content
AI-generated narration uses a synthetic voice to produce speech from text. It can be a practical option for training that requires multiple language versions or recurring revisions. The decision should depend on an approved sample, however, rather than a generic voice demonstration. A pleasant greeting tells you little about how the voice handles your terminology or a dense explanation.
Ask which voices and language variants are supported for the project. Test a representative passage containing acronyms, numbers, product names, and transitions between sentences. Listen for clarity, emphasis, pauses, and consistency across the passage. Approve the delivery in the context of the video, where the available time may differ from an unrestricted audio sample.
Synthetic narration may allow revisions without booking another recording session, but revised output still needs checking and integration. A human-recorded option, where available, may offer a different kind of expressive direction. Neither is automatically suitable for every lesson. Confirm whether such alternatives are actually offered and what their production and revision arrangements would be.
Do not assume that approval of one language establishes suitability in another. Regional pronunciation preferences, borrowed terminology, and the available voice choices can differ. Also confirm any restrictions relevant to the intended distribution. If the project requests a particular person’s likeness or voice, treat authorization and availability as separate questions rather than assuming standard narration covers them.
Approve Pronunciation Separately from Spelling
Build a short pronunciation reference before producing the full set. Include names, abbreviations, product codes, and terms that have an established internal pronunciation. State whether an acronym should be spoken as a word, expanded, or read letter by letter. An approved spelling does not settle how every voice should say it.
Give the language reviewer context and, where helpful, an authorized reference recording. The reviewer should listen to the generated result, not only inspect the script. Check numbers and units in sentences, especially where an abbreviation has more than one possible reading. Keep technical terms consistent with the course’s other materials.
Some production methods use pronunciation instructions or altered input spellings to influence speech. Keep those technical inputs separate from the approved display text. A spelling used to make a voice pronounce a name correctly should not accidentally enter the captions, transcript, or terminology glossary as the official written form.
Document approved pronunciations and voice settings for later revisions. When a segment is regenerated, listen to the adjoining segments too: pacing or tone may change at the edit. A pronunciation approval is most useful when tied to the actual reviewed audio, language variant, and script revision.
Give Timing and Text Their Own Review
Timing is more than making the audio end when the video ends. Explanations should accompany the relevant scene, and learners need time to follow the action. Review pauses, scene transitions, and the points where a visual instruction depends on a spoken phrase. Agree whether the video can be extended, recut, or divided into shorter segments.
For text, set project conventions for cue duration, line breaks, placement, and reading pace in the intended player. Apply them with linguistic judgment. Avoid splitting a tightly connected phrase unnecessarily, and check that a cue remains visible long enough to understand. There is no useful single timing promise for every language, audience, and screen layout.
Automatic transcription or caption generation can assist preparation, but its output needs verification. W3C cautions that automatic captions must be checked for accuracy. Review against the final soundtrack, including any required speaker or sound information. Then inspect playback: a correct text file can still be out of sync with the delivered edit.
Keep edits traceable. If the translated narration changes after text approval, identify the affected cues and transcript passages. If a video scene moves, check timestamps throughout the affected section. Approving each asset separately is not enough when the relationship between them has changed.
Test the Sample in the Intended Player
Before scaling production, review a short but representative sample in the learning platform or player that learners will use. Include a difficult term, a busy visual, and a scene transition. Ask reviewers to assess meaning, listening clarity, reading time, and whether the relevant action is still visible. Record specific corrections rather than a general preference for one voice.
Confirm how language tracks are selected and whether captions can be turned on and off. W3C’s media-player guidance also addresses keyboard operation and clearly identified controls. Assign player configuration and accessibility testing to the responsible platform team; supplying a caption file does not configure the player automatically.
W3C guidance on accessible media players
Agree the required file format with that team. A separate timed text file, such as SRT or WebVTT, offers a different delivery arrangement from text permanently embedded in the video. Check the actual platform’s support, including language labels and character display. If both downloadable and streamed versions will be distributed, test both rather than assuming they behave identically.
Plan Updates as a Set of Related Assets
Give each approved script segment a stable identifier and track its related audio, text cues, and video location. A changed sentence can affect all of them. Keep source and target-language versions associated with the same content revision so that a new caption file is not accidentally paired with an old soundtrack.
For frequent updates, ask whether narration can be delivered in segments as well as in the final mix, and whether editable timed text is included. Identify who will retain the production files and voice settings. A finished MP4 is useful for viewing, but it is not a complete maintenance package for every kind of future change.
At delivery, include a manifest that identifies the language, script revision, approved voice reference, and intended use of each file. Open the package as a recipient would and confirm that the correct audio and text accompany the correct video. This small handoff check can expose naming or packaging mistakes that a review of the individual assets would miss.
Specify review responsibilities in the proposal. Linguists assess language and pronunciation; the content owner approves technical meaning and instructional changes; multimedia specialists implement synchronization and presentation; the platform team checks deployment. These are recommended responsibilities to agree for the project, not a claim that every engagement includes every role or review round.
Effort depends on more than runtime. Missing scripts, inseparable audio, dense terminology, unsupported delivery formats, and late changes can all affect production. Ask the estimate to identify translation, narration, timed text, adaptation, testing, and revision scope. Compare supported options against the same sample and deliverables rather than comparing headline prices alone.
Discuss Supported Options with Terralíngua
Terralíngua advertises multimedia localization, AI-generated narration, and video localization including subtitles. Its multimedia page describes linguistic review of synthetic pronunciation. These services can support translated training, but the available voice, language variant, and delivery method should be confirmed for the specific project.
Share a sample video and language list, then discuss the script, pronunciation needs, screen text, and required outputs. Ask what can be retained from the source and what needs adaptation. Human-recorded narration, specialist accessibility deliverables, and particular voice choices should be confirmed rather than assumed to be included.
Explore Terralíngua’s multimedia localization and AI narration services
Read about Terralíngua’s video localization and subtitle services
Frequently Asked Questions
Can We Start with Subtitles and Add Narration Later?
Yes, this can be planned as a staged project. Retain the approved translation and editable text files, but allow for adaptation to spoken delivery and synchronization when narration is added. Subtitles timed to the original speaker will not necessarily match the new audio. Confirm which assets can be reused before budgeting the second stage.
Should the Text Match the Translated Voice Word for Word?
First define the text’s purpose. Captions should accurately represent the final audio and required sound information, while translated subtitles may be adapted for reading. Agree the editorial convention with the reviewer and accessibility owner. Avoid untracked differences that change an instruction, terminology, or qualification between what learners hear and read.
What If a Preferred Voice Cannot Pronounce a Key Term?
Request a reviewed correction sample using the available pronunciation controls. If the result remains unsuitable, discuss another supported voice or production approach. Do not approve an incorrect term because the rest of the voice sounds natural. Confirm any alternative’s availability before making it part of the project plan.
Can a Text Correction Be Made without Exporting the Video Again?
Sometimes. A separately loaded subtitle or caption track may be replaceable without changing the video, depending on the platform. Text embedded in the picture requires a new video export. A correction that changes spoken meaning or timing also needs review of the audio and related assets, even if the text file itself is easy to replace.
Send a Sample That Reveals the Real Choices
To request a useful assessment, send Terralíngua a representative video, the available script and source files, target languages and variants, audience details, intended player, and deadline. Include difficult pronunciations and any accessibility specifications. Ask for a supported-options discussion covering narration, subtitles, review, and multimedia adaptation before selecting the final production approach.
Request a quote for multilingual training narration and subtitles