Localized interface label narration must match the product that the learner actually sees. A correct translation of an English button is not necessarily the label used in the localized application. Review the interface, script, and screenshots together, then approve the spoken instruction against the real task before generating the full tutorial.
Capture the target interface state
Record the application's version, selected interface language, account role, and relevant settings. A label can differ because of a product update or permission level rather than a translation mistake.
Capture each screen used in the tutorial and give it a stable ID. Pair that ID with the action and expected result. Avoid relying on a screenshot filename that describes only its creation date.
Create a terminology list from the actual interface: page titles, menu entries, buttons, field names, and status labels. Mark labels that remain in the source language or differ between desktop and mobile layouts.
Use Microsoft's localization style-guide resources as general language context where relevant. The product's actual approved labels remain the source for instructions about that product.
Link each spoken action to a visible control
Write the instruction with the exact localized label and enough context to find it. If two controls share a label, identify the section or purpose. “Select Save in the profile panel” is more useful than an isolated “Save.”
Do not translate a label again for stylistic elegance. The learner needs to recognize the string on screen, even if another wording would sound more natural in ordinary prose. Keep explanatory language around the label natural.
Follow the step-by-step software audio pattern: starting location, action, and confirmation. The confirmation must also match the localized interface's actual result.
If the application itself is not localized, decide how the tutorial will refer to its source-language labels. An approved explanation can bridge the language difference, but the narration should not invent a translated control that does not exist on screen.
Review text, pronunciation, and timing separately
First perform bilingual script QA to check that the instructions preserve the correct action and conditions. Then have the locale reviewer read the target script against the screenshots.
Test label pronunciation in context. A product may combine a local sentence with an untranslated English feature name or an acronym. Keep the normal visible label in the transcript while documenting any production spelling used for audio.
Measure the time needed to locate the control and see the result. A translated label may be longer, but the practical issue is whether the learner can follow the action. Extend the screen recording or split the step where appropriate.
Record any deliberate difference between the visible string and its spoken form, such as expanding an abbreviation. A future reviewer should be able to distinguish an approved explanation from an accidental mismatch.
Verify the localized task from start to finish
Use the intended interface language and account role to perform the whole task while listening to the assembled tutorial. Do not approve solely from the script sheet. Menus may appear in a different order or a control may move at the screen width used in the final recording.
Check the first and last words of each segment, especially where a label ends near a visual transition. A correct label that is cut off by the next screen is still unusable instruction.
Track the interface version, script revision, audio revision, and screenshot set in a multilingual version matrix. When the product changes, that record identifies which tutorials need review.
Generate one approved task segment in the voice studio, download it, and test it against the real interface. Continue the batch only when a reviewer can locate the named control, perform the stated action, and observe the promised result without guessing.
