Translate Audio Description Scripts Correctly

How do you translate audio description scripts, timed visual cues and dialogue gaps correctly? Start by identifying exactly what the source record tells the reader to understand or do, then preserve its names, numbers, conditions and limitations. This guide provides a practical, step-by-step method for audio description translation and accessible video localization, with worked examples, common mistakes and quality-control checks for readers anywhere in the world.

People searching for how to translate audio description, AD script localization, video description narration, dialogue gaps, visual cues, accessibility WCAG and extended audio description need more than a word-for-word translation. They need a reliable explanation of the specialist document’s structure. A subtitle translates speech into text; audio description makes necessary visual information audible. Confusing the two produces the wrong deliverable. Even a beautifully translated description can become unusable if it is spoken over a character’s crucial line or if a visual action is named before it happens. The solution is to separate verified content from background explanation, then confirm that the target-language reader makes the same practical interpretation as someone reading the source.

Quick summary: Help blind and low-vision audiences receive the essential visual information of a film or learning video in the target language without masking significant dialogue, sound or on-screen facts. The method is to identify the source authority, lock identifiers and units, explain domain terms, retain every condition or uncertainty, produce natural target-language prose and test the result against the original. Accessibility evaluation depends on the complete experience and applicable content standard. A script alone is not proof that the final mixed video meets accessibility requirements; timing, recording, mix and user testing matter.

The problem: a correct-looking sentence can still be wrong

A subtitle translates speech into text; audio description makes necessary visual information audible. Confusing the two produces the wrong deliverable. Even a beautifully translated description can become unusable if it is spoken over a character’s crucial line or if a visual action is named before it happens. Audio description is timed editorial narration, not exhaustive commentary on everything in the frame. The translator must preserve essential information while working within the actual space available in the soundtrack.

Translation errors are often invisible because a finished sentence sounds convincing. The reader does not know which qualifier was removed, which technical label was replaced or which measurement was detached from its field. A serious educational guide therefore treats source structure as evidence: what is said, how it is supported, and what it does not say. Rather than promising an answer from vocabulary alone, it teaches a chain of checks that a learner can repeat.

Definition: what this translation task includes and does not include

audio description translation and accessible video localization is the act of carrying documented information about audio description scripts, timed visual cues and dialogue gaps into another language without changing what the document establishes. Its scope includes labels, explanatory notes, warnings, dates, figures, symbols, relationships, qualifiers and source-specific conventions. The translator may improve readability or add a clearly marked explanation; the translator may not silently create missing facts, change a status, or assert a result that the original document never provided.

Audio description is timed editorial narration, not exhaustive commentary on everything in the frame. The translator must preserve essential information while working within the actual space available in the soundtrack. This is why a bilingual dictionary is only one tool. It can suggest vocabulary but cannot confirm which identifier belongs to which object, whether a planned action actually happened, whether a condition is currently valid, or whether an abbreviated table legend changes the meaning. Those decisions require context and disciplined checking.

The canonical job of this article is help blind and low-vision audiences receive the essential visual information of a film or learning video in the target language without masking significant dialogue, sound or on-screen facts. It does not replace the broader audio description translation and accessible video localization standards or an expert review of a live, safety-sensitive decision. For general meaning extraction, use the separate eduKateSG guide on translating meaning rather than words. For style, idioms and systematic terminology checks, follow the linked specialist guides at the end instead of making this article compete with those broad teaching routes.

The first-principles foundation

1. Description is not subtitling

Subtitles generally show speech or relevant sound as text. Audio description adds voiced information about otherwise inaccessible visual facts. Those tasks can complement each other but cannot be treated as substitutes. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

A practical way to teach this principle is to underline the source phrase, name its function in plain language, and write a test question before drafting. For audio description translation and accessible video localization, ask what observable fact, relation, category or instruction would be misread if this phrase were missing. A translation is ready only when the target answers that test question without requiring the reader to reconstruct the original language.

2. Meaningful visuals take priority

An action that changes a character’s choice or reveals the next step in a demonstration is more important than decorative scenery. Translate necessary content before atmosphere. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

3. Pauses are real constraints

Narration usually fits existing gaps in dialogue. Target languages may expand in length, so accurate sentences still require timing and performance review. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

4. Not every silence is empty

A quiet beat may contain decisive music, a sound effect or deliberate dramatic silence. Dialogue gaps must be checked against the full soundtrack, not just the transcript. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

5. Character reference must be anchored

A visual person should be identified by established descriptors or names available to the audience. The translator should not invent identity, relationships or motives. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

6. Description is observational

‘She turns away and closes the notebook’ is a visible action; ‘she is ashamed’ is an interpretation unless the work clearly establishes it. Preserve the source’s evidential level. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

7. Visual information has timing

A gesture, title card or sudden reveal belongs at a particular moment. If the target narration lands early, it may spoil the meaning of the scene. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

8. On-screen words can matter

An essential warning, instruction, chart label or text message may be inaccessible through sound. Translate it when needed and make its location in the scene understandable. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

9. Sound may already explain action

A door slamming or a name spoken aloud can convey facts without additional narration. Added description should complement, not redundantly overwhelm, the existing track. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

10. Extended description is distinct

Where pauses are insufficient, an extended audio-described version may stop the video for additional narration. Do not assume a standard track can simply run longer over dialogue. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

11. Language choices affect fit

Target-language grammar, word order and proper names influence syllables and speaking pace. The right solution is sometimes concise rephrasing, not speed-reading. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

12. Final accessibility is tested

Review the audio alongside visuals and any original track. A text document with no timing review cannot establish that important information is actually accessible. This distinction determines what the target sentence is allowed to claim. When a reader can identify the original evidence and the relevant limitation, the translation is useful; when a translator erases the distinction in pursuit of elegant wording, fluency can become camouflage for an error.

A repeatable twelve-stage translation method

Apply these stages in order on a short example before attempting a full document. Different stages prevent different kinds of error. Source identification protects provenance; the numerical pass protects measurements; the reader test protects the real-world meaning of the finished text. Separating these jobs is more reliable than trying to translate and approve everything in one hurried read.

Step 1: Identify the media and audience

Determine whether the target is fiction, documentary, an educational demonstration, a training module or a short social video. The information priorities differ. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Use a two-column comparison: original evidence on the left and the proposed target expression on the right. For audio description scripts, timed visual cues and dialogue gaps, include an additional column for what must remain fixed. A reviewer should be able to check the relationship, not merely see that every original word appears somewhere. Mark each item checked, uncertain or returned to the document owner; never treat uncertainty as a completed pass.

Step 2: Watch before translating

View the entire scene with audio and visuals. A script segment alone may misidentify a character, action or clue whose meaning becomes clear later. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 3: Inventory missing visual information

List key actions, people, locations, changes and on-screen text that the existing soundtrack does not communicate. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 4: Read the timecoded script

Check in and out points, scene cuts, pause lengths, character names and annotations. A line is a timed cue, not a standalone essay sentence. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 5: Define description priorities

Rank plot-essential or task-essential facts ahead of incidental colour and decoration. Preserve intentional reveals and causal relations. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 6: Draft accurate target wording

Create clear target-language sentences that sound natural when spoken aloud. Do not add unstated motives or visual details absent from the image. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 7: Estimate spoken duration

Read the line at a comfortable narration pace inside its cue. Compare with available gap rather than applying a universal words-per-second formula blindly. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 8: Respect the soundtrack

Check dialogue, sound effects, music and meaningful silence. Avoid placing narration where listeners need original audio to understand the scene. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 9: Resolve overflow transparently

Shorten without dropping key meaning, move to a valid nearby gap if editorially appropriate, or escalate for extended description rather than talking over content. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 10: Test speaker continuity

Listen with the picture covered. The audience should know who is acting, when a scene changes and what an ambiguous pronoun refers to. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 11: Check accessibility with users

Where possible, include qualified blind or low-vision reviewers and speakers of the target language in the review loop. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Step 12: Log and version the mix

Keep script timing, narration recording, edit history and final mix specification together so later revisions do not undo accessibility work. Write a short check note beside the source, not inside the translation itself, so that the public text can remain readable. Keep any unresolved source ambiguity visible in the review log instead of settling it by intuition.

Worked translation examples: preserve the evidence

The following mini-cases are illustrative teaching scenarios, not live datasets or instructions to rely on without consulting the actual source. In each case, identify the apparent translation, the invisible error, the corrected reasoning and the practical check. The technique transfers between languages because it focuses on relationships and evidence rather than one pair of dictionary equivalents.

Worked example 1: Character introduction

Illustrative source situation. A scene shows a woman in a red coat entering silently. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. A translation that calls her ‘the detective’ before that role is established invents plot information. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Describe the visible coat and entrance using the established character reference. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Listen without picture: is identification clear without premature revelation? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 2: Visual reveal

Illustrative source situation. A character opens a box; the camera reveals an empty compartment. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Narrating ‘the box is empty’ before the lid opens spoils the visual beat. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Keep the description in the available gap after the reveal. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Check timing against the edit and dialogue. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 3: On-screen warning

Illustrative source situation. A training video displays ‘Emergency stop’ while the instructor is silent. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Omitting the text makes a crucial control inaccessible. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Give an audible equivalent of the visible label at the correct moment. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Could the learner identify the correct control without seeing the screen? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 4: Dialogue overlap

Illustrative source situation. A timed cue is longer in the target language than in the source. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Reading quickly over dialogue may conceal a decisive spoken line. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Restructure the sentence to fit or flag the cue for editorial retiming. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Audition the actual narration in the complete mix. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 5: Silent gesture

Illustrative source situation. A student shakes her head after the teacher offers an answer. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Calling it agreement reverses the action. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Render the negative head movement clearly without inventing a reason. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Can a listener tell that the offer was rejected? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 6: Descriptive inference

Illustrative source situation. A person looks at a letter and begins to cry. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Adding ‘because her father has died’ invents backstory not visible or spoken. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Describe the reading and tears; let the story supply cause when it becomes known. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Separate physical evidence from interpreted motive. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 7: Scene change

Illustrative source situation. The scene cuts from an office to a coastal road. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Dropping the location transition makes the next voice sound as if still in the office. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Use a brief location cue at the cut where audio allows. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Check the listener’s mental map. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 8: Sustained silence

Illustrative source situation. A long pause contains a distinctive alarm tone. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Filling the gap with decorative description can mask the plot-significant alarm. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Keep the audible cue available and shorten or move incidental description. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Listen for information already carried by sound. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 9: Screen message

Illustrative source situation. A phone screen reads ‘Train cancelled’ while a character says nothing. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Translating the moment as ‘she reads a message’ misses the actionable content. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Include the message meaning and identify it as on-screen text. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Does the listener know what changed? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 10: Printed chart

Illustrative source situation. A video displays a rising line graph, while narration says ‘look at this’. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. A vague AD line ‘a graph appears’ does not convey the result. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Describe the direction, axis meaning and important change using visible values supported by the chart. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Can a learner explain the conclusion without the picture? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 11: Button location

Illustrative source situation. A tutorial silently highlights a button at the bottom left of an interface. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Replacing precise position with ‘click here’ fails non-visual comprehension. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Describe the functional button label and navigational position where relevant. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Does the instruction remain actionable when the screen is hidden? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 12: Colour relevance

Illustrative source situation. Two wires are distinguished only by red and blue in a repair video. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Eliminating colours because they seem decorative removes the distinction. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. State the colour differences and any printed labels without inventing a safer wiring procedure. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Does the listener have enough information to identify the referenced item? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 13: Unimportant colour

Illustrative source situation. A character’s chair is pale yellow but its colour never matters. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. A long phrase about the chair can crowd out a plot action. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Omit low-priority details when the cue is tight; preserve the consequential movement. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Review what the listener must know next. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 14: Montage

Illustrative source situation. A rapid sequence shows three different workplaces. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. One continuous description can imply everything occurs in a single location. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Use concise, ordered scene cues that respect cuts. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Can the listener follow the sequence from first to last? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 15: Flashback

Illustrative source situation. Visual desaturation marks a flashback that the dialogue never announces. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Ignoring the temporal transition can confuse chronology. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Signal the shift when supported by editing context without over-specifying an unknown year. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Compare target chronology with the film’s actual cues. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 16: Speaker identity

Illustrative source situation. An off-screen voice speaks while the camera shows a different character. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Describing the visible person as the speaker misattributes dialogue. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Keep visual action and audio speaker separate unless the source establishes identity. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Can a reader identify who is visible and who is heard? Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 17: Subtitle mismatch

Illustrative source situation. A team supplies an SRT subtitle file as if it were audio description. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Text dialogue captions alone do not convey silent actions. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Create or request a dedicated AD script and track; do not label subtitles as AD. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Check the deliverable includes voiced visual information. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 18: Original sound effect

Illustrative source situation. A glass breaks audibly and is also shown on screen. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Repeating ‘a glass breaks’ can be redundant if the sound already explains the event. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Prioritise any essential unseen aftermath instead. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Listen to the track with eyes closed before writing. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 19: Visual joke

Illustrative source situation. A character proudly displays a sign that accidentally reads the opposite of the intended message. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. A description of ‘a funny sign’ hides the joke. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Convey the sign’s wording and the visible contrast at the moment the audience can understand it. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Avoid explaining emotions that viewers are expected to infer. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 20: Proper name

Illustrative source situation. A name badge is visible but not mentioned in dialogue. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Guessing the person’s role from appearance can create false identity. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Voice the badge text if it is story-relevant, with an audible introduction. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Confirm the spelling against the graphic rather than memory. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 21: Text direction

Illustrative source situation. A target script uses a right-to-left language, but the visual card remains left-to-right. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Reordering the actual screen elements in the description can misidentify their positions. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Describe source-screen positions accurately while drafting natural target-language wording. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Check referenced objects against the actual frame. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 22: Technical overlay

Illustrative source situation. An instructional video has a small ’12 V’ label next to equipment. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Dropping the unit turns the label into an ambiguous number. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Read the value and unit as the visible technical annotation. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Keep the same object-value association. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 23: Unclear pronoun

Illustrative source situation. ‘She hands it to her’ follows shots of two women. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Repeating the source pronouns in a target language may conceal who gives and receives. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Use established identifiers when necessary and supported by the visual evidence. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Listen without picture for referent confusion. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 24: Reaction shot

Illustrative source situation. A director holds on a character’s surprised expression after news arrives. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Describing it before the spoken news can spoil the response. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Time the reaction phrase after the line so the response stays causally legible. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Check the reveal order. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 25: Narration speed

Illustrative source situation. A 16-word target line fits a 1.5-second gap only when rushed. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Rushed delivery is not a valid accessibility solution simply because it is technically possible. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Shorten responsibly or request extended description as appropriate. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Listen at normal speed and test comprehension. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 26: Music cue

Illustrative source situation. A distinctive motif returns over a silent transition. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. A description that covers the motif may erase a narrative connection. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Keep essential musical information audible and relocate less important visual description. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Test both plot information channels. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 27: Extended description

Illustrative source situation. A dense science animation conveys six steps during uninterrupted speech. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Trying to speak every step over the lecturer loses both lecture and visual meaning. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Consider an explicitly extended audio-described version that pauses playback if the programme permits it. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Confirm whether the format and accessibility target allow the additional track. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 28: Live versus prerecorded

Illustrative source situation. A production team requests live commentary but supplies a prerecorded AD script. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. Assuming the same delivery workflow may cause missed timing and handoff requirements. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Identify and label the deliverable, then adapt rehearsal and cue control to that format. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. The reader should know what will actually be performed. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Worked example 29: Accessibility audit

Illustrative source situation. A document declares ‘accessible’ solely because a script exists. Begin by identifying the kind of document this statement appears in, the relevant field or visual label, and the scope of its claim. Source context matters because the same everyday expression can stand for a controlled technical category, an instruction, a measurement or a report of what has already occurred.

The tempting error. An unrecorded line cannot demonstrate that the final media meets the intended standard. This may sound fluent in English or another target language, but the reader would reach a different conclusion. It is a semantic and functional error rather than a minor preference about style. A good reviewer explains the changed relationship precisely instead of marking the sentence merely ‘awkward’.

Corrective translation decision. Describe the script as ready for timed recording and subsequent mix and user checks. Choose natural target-language grammar only after these facts are fixed. If a short term needs explanation, place the explanation beside the source identifier or label; do not overwrite the identifier with a guess. If the evidence is insufficient, retain the uncertainty and request clarification rather than inventing a complete answer.

Reader test. Verify the playable output, not only the text file. Have another person answer this question from the target alone. Compare their answer with the source’s actual meaning, including any negative statement, condition, date, unit or location. If the two interpretations differ, revise the translation and test again rather than assuming that polished vocabulary proves correctness.

Failure modes and how to repair them

These are the errors that survive ordinary spelling checks. A clean document can still contain reversed relationships, invented certainty or missing scope. Use this section as a specialised prepublication audit and require a documented correction when any failure is found.

Failure 1: Calling captions audio description

A subtitle file can preserve dialogue but not narrate visual facts. Keep the media type and accessibility purpose distinct. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

An effective quality-control record contains the faulty phrase, the corrected version, the source evidence and the reason the change matters to the reader. That brief record makes repeat mistakes less likely when the same type of document appears again, especially when several translators or editors share the work.

Failure 2: Overwriting dialogue

A longer translated phrase may obscure essential speech when recorded. Test the real target-language performance in its actual cue. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 3: Describing every object

Exhaustive inventories overwhelm the listener and may omit the actual plot or learning task. Prioritise consequences and actions. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 4: Inferring feelings and motives

A frown may be visible; a hidden motive usually is not. Avoid turning visual evidence into confident psychology. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 5: Spoiling visual revelations

Narration timed ahead of the image can announce a discovery too early. Compare cue in-points against the edit. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 6: Ignoring sounds

A sound effect, musical cue or meaningful silence can communicate information that added narration should not mask. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 7: Confusing speakers

An off-camera line is not necessarily spoken by the person in the frame. Keep speaker labels and visual descriptions separate. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 8: Dropping on-screen text

Safety warnings, chart axes and interface labels may be essential even if short. Translate what a listener needs to operate or understand. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 9: Using vague deictics

‘This’, ‘there’ and ‘over here’ may refer to invisible locations. Use functional names and clear referents. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 10: Forcing identical sentence lengths

Languages expand and contract. Timing must be reviewed after meaning, not imposed mechanically on every source word. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 11: Claiming compliance too early

A good script is only one component; the recorded mix and delivered player experience require testing. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Failure 12: Ignoring updates

Editors may recut scenes after translation. An old cue can narrate the wrong event when timing changes. To repair the problem, return to the smallest verifiable unit in the source and compare its field, neighbour, caption or footnote. Then rewrite only what the source justifies. Keep the corrected wording understandable to non-specialists without hiding the controlled technical meaning.

Practice and transfer: do the translation thinking yourself

These exercises deliberately provide an imperfect rendering. Before reading the recommended repair, identify what evidence has changed and whose action or understanding could be affected. Repeat the exercise a few days later with the answer hidden. This turns a checklist into a habit that can transfer to unseen documents.

Practice 1: A script says ‘she is furious’ when the video on…

Task. A script says ‘she is furious’ when the video only shows a woman turning away. First identify the source fact or relation that must remain stable. Write a candidate translation in two sentences: one with the essential information and one with any necessary clarification. Do not introduce a claim absent from the prompt.

Reasoned repair. Rewrite the line to describe the visible action without assigning a mental state. After correcting the error, explain aloud why the original would have misled a reader. Finally, design a one-line quality check that would catch the same error in an unfamiliar document. If that check depends on a fact not supplied, state which source you would consult.

Practice 2: An important in-video warning appears for four s…

Task. An important in-video warning appears for four seconds and nobody reads it aloud. First identify the source fact or relation that must remain stable. Write a candidate translation in two sentences: one with the essential information and one with any necessary clarification. Do not introduce a claim absent from the prompt.

Reasoned repair. Write concise, audibly intelligible wording and propose a cue that avoids the existing dialogue. After correcting the error, explain aloud why the original would have misled a reader. Finally, design a one-line quality check that would catch the same error in an unfamiliar document. If that check depends on a fact not supplied, state which source you would consult.

Practice 3: Two characters exchange a note while an unseen p…

Task. Two characters exchange a note while an unseen person speaks. First identify the source fact or relation that must remain stable. Write a candidate translation in two sentences: one with the essential information and one with any necessary clarification. Do not introduce a claim absent from the prompt.

Reasoned repair. Keep the visual handoff and the off-screen voice distinct; do not claim the visible person is the speaker. After correcting the error, explain aloud why the original would have misled a reader. Finally, design a one-line quality check that would catch the same error in an unfamiliar document. If that check depends on a fact not supplied, state which source you would consult.

Practice 4: A translated sentence takes five seconds but the…

Task. A translated sentence takes five seconds but the dialogue gap lasts three. First identify the source fact or relation that must remain stable. Write a candidate translation in two sentences: one with the essential information and one with any necessary clarification. Do not introduce a claim absent from the prompt.

Reasoned repair. Identify what is essential, edit for brevity or flag a suitable alternative rather than recording over speech. After correcting the error, explain aloud why the original would have misled a reader. Finally, design a one-line quality check that would catch the same error in an unfamiliar document. If that check depends on a fact not supplied, state which source you would consult.

Practice 5: A documentary’s historical date appears silently…

Task. A documentary’s historical date appears silently on a title card. First identify the source fact or relation that must remain stable. Write a candidate translation in two sentences: one with the essential information and one with any necessary clarification. Do not introduce a claim absent from the prompt.

Reasoned repair. Preserve the date and specify that it is an on-screen title, without inventing surrounding historical explanation. After correcting the error, explain aloud why the original would have misled a reader. Finally, design a one-line quality check that would catch the same error in an unfamiliar document. If that check depends on a fact not supplied, state which source you would consult.

Practice 6: A rising bar chart is described as ‘a chart with…

Task. A rising bar chart is described as ‘a chart with numbers’. First identify the source fact or relation that must remain stable. Write a candidate translation in two sentences: one with the essential information and one with any necessary clarification. Do not introduce a claim absent from the prompt.

Reasoned repair. State the data relationship the visible chart shows and retain the correct units and axes. After correcting the error, explain aloud why the original would have misled a reader. Finally, design a one-line quality check that would catch the same error in an unfamiliar document. If that check depends on a fact not supplied, state which source you would consult.

Practice 7: A character’s name is revealed on a badge near t…

Task. A character’s name is revealed on a badge near the end of a scene. First identify the source fact or relation that must remain stable. Write a candidate translation in two sentences: one with the essential information and one with any necessary clarification. Do not introduce a claim absent from the prompt.

Reasoned repair. Time the name so it is not spoken before its onscreen revelation. After correcting the error, explain aloud why the original would have misled a reader. Finally, design a one-line quality check that would catch the same error in an unfamiliar document. If that check depends on a fact not supplied, state which source you would consult.

Practice 8: An AD script is complete but the audio track has…

Task. An AD script is complete but the audio track has not been mixed. First identify the source fact or relation that must remain stable. Write a candidate translation in two sentences: one with the essential information and one with any necessary clarification. Do not introduce a claim absent from the prompt.

Reasoned repair. Describe the deliverable as a completed script requiring performance and technical review, not a finished accessible video. After correcting the error, explain aloud why the original would have misled a reader. Finally, design a one-line quality check that would catch the same error in an unfamiliar document. If that check depends on a fact not supplied, state which source you would consult.

Four-week learning plan for independent translators

Week one: read and classify

Collect ten real, publicly available examples with permission to reuse short excerpts. Mark each identifier, source date, measured value, label, qualifier and instruction. Do not translate until you can explain the records in plain language. Correctly identifying the unit of meaning is the foundation for every later step. Record what improved and what remains uncertain. A translator who can explain the repair process is better prepared for a new source than a translator who only keeps a polished final page.

Week two: make controlled drafts

Translate five small documents or excerpts, one stage of the method at a time. Maintain a glossary with a source term, its document-specific definition, the preferred target rendering and one rejected misleading alternative. Compare every figure and status against the originals before editing for fluency. Record what improved and what remains uncertain. A translator who can explain the repair process is better prepared for a new source than a translator who only keeps a polished final page.

Week three: challenge the translation

Exchange drafts with a partner or perform a delayed self-review. Hide the originals and ask what a reader would do or conclude. Then reveal the originals and list mismatches by severity: identity, meaning, safety, timing, uncertainty, style and readability. Repair the highest-consequence errors first. Record what improved and what remains uncertain. A translator who can explain the repair process is better prepared for a new source than a translator who only keeps a polished final page.

Week four: transfer into a new context

Translate a new subgenre within the same field that uses unfamiliar layouts. Build a fresh source map, verify its terms and produce a short editorial decision log. The objective is not to memorise these examples but to reproduce the checking method when vocabulary and document shape change. Record what improved and what remains uncertain. A translator who can explain the repair process is better prepared for a new source than a translator who only keeps a polished final page.

Frequently asked questions

What is audio description?

It is spoken narration of significant visual information that the existing soundtrack does not adequately convey. It supports people who cannot see or interpret the visual content. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

Is audio description the same as subtitles?

No. Subtitles display speech and relevant sound as text, while audio description adds an auditory channel for essential visual information. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

Why translate AD separately from dialogue?

AD has its own timing, audience purpose and editorial constraints. A sentence that works on paper may not fit the soundtrack or identify the right event. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

How much visual detail belongs in a cue?

Enough to communicate the information needed to understand the scene or task, without drowning out dialogue or conveying unsupported interpretation. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

Can AD overlap with dialogue?

Standard description typically uses suitable pauses; overlapping significant speech can hide meaning. Editorial alternatives may include concise phrasing or an extended version. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

What is extended audio description?

It is an approach that pauses the video so important visual information can be narrated when normal dialogue gaps are insufficient. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

Does every silent moment need narration?

No. Silence, music and sound effects can themselves convey information. Describe the visual content that is otherwise missing and important. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

May I identify a character by profession?

Only when the film or supporting source establishes that role. Visible clothing, name badges and established character references are safer than guesses. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

Should I read important on-screen text?

When the words carry information necessary to understand or act, provide an audible equivalent, accounting for timing and clarity. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

Can I automatically translate AD using machine translation?

Automated drafts can help, but timing, pronoun reference, scene evidence and listenability require editorial checking against the actual video. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

How do I test a translated AD track?

Listen to the final mix without watching the picture, then cross-check content against the visuals and obtain relevant accessibility review when possible. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

Does a finished script prove WCAG compliance?

No. The final media delivery must provide the needed audio description in an accessible experience; a script is one part of that process. For a real document, apply this answer in conjunction with its own source definitions and current authoritative guidance rather than assuming every publisher uses identical conventions.

Quality-control checklist before publication

Every identifier, place, person, item or route is matched to the same source record.
Every number retains its unit, sign, denominator, time convention and relevant footnote.
Observed facts, estimates, predictions, conditions and intended future actions remain distinct.
Missing information remains missing instead of being completed by a translator’s guess.
Safety, access and consent language has been compared separately against the source.
Labels remain attached to the correct diagram, table row, cue or narrative step.
All interpretation changes are supported by the named primary source or documented author clarification.
A target-language reviewer can correctly explain the reader’s practical takeaway.
The translation includes source date and edition when these affect interpretation.
The final content has been checked for natural, readable English without losing technical meaning.

Selected primary references and specialist learning routes

The external references below provide domain definitions or authoritative safety and accessibility principles. They are used as sources for learning the terminology and checking the scope of particular claims, not as a licence to reuse a standard or provide an independent technical certification.

For translation skills shared across documents, continue through these eduKateSG routes. The first four own broad translation principles; this specialist article owns only the particular evidence and reader tasks explained above.

The practical conclusion is simple: the best audio description translation and accessible video localization work preserves the evidence before polishing the language. A confident translator can point to the original identity, measurement, condition, action and uncertainty behind every consequential sentence. That is how translation becomes useful to a real reader rather than merely correct-looking text.

Top Ways to Translate Correctly | Translate Audio Description Scripts, Visual Cues and Dialogue Gaps Without Losing Accessibility