How do you translate “this one,” “the one over there,” or “move it here” when a speaker points at something in a video? The words alone are often insufficient. Accurate translation of pointing and screen references requires matching the spoken phrase to the visible object, time, camera view, and task before selecting a natural target-language expression. Otherwise a grammatically correct subtitle can direct the audience to the wrong item.
People searching for how to translate “this one” in a video, how to translate pointing gestures in subtitles, how to translate “here” and “there” with visual context, or how to translate a screen recording accurately are dealing with referential language. Expressions such as this, that, here, there, these, those, the left one, above and below do not identify their referents by themselves. Their meaning depends on what the speaker and viewer can see.
This guide concentrates on spoken references anchored in a visual scene. It is not a dictionary of gesture meanings; it explains how to keep words, images, timing and action aligned across languages. It applies to product demonstrations, lessons, tutorials, training videos, crafts, cooking instructions, software screen recordings, assembly guides and ordinary conversation on camera. Start with eduKateSG’s five-layer translation method for the general meaning process; the special challenge here is locating what a phrase points to before translating it.
Quick answer: find the referent before translating the word
Pause at the spoken reference and ask five questions: what expression was said, where was the speaker looking or pointing, which visible object or location was available, what action follows, and how does the target language normally identify that referent? Only after answering should you write the subtitle or voiceover. “This” might refer to an ingredient, a button, a process on the screen, the current step, or an entire previous explanation.
A practical rule is to track each referring expression to a visual anchor. If the speaker says, “Click this one,” and the cursor highlights the lower of two identical buttons, the translation must preserve that choice. A version that simply says “click the button” may be sufficient if one button is visually singled out. A version that says “click the top button” is false if the cursor points to the lower one.
When evidence is weak, avoid invented specificity. A translator may add a clarifying noun, such as “this icon,” only when the icon is actually identifiable. It is better to preserve a source ambiguity or request the missing frame than to make a subtitle confidently name the wrong object.
The real problem is reference, not the dictionary
Deictic expressions are words whose meaning depends on a point of view or context. “Here” refers to a location relative to someone or something; “that one” selects an item from a set; “now” can refer to a moment in the recording or in an instructed procedure. Languages divide this territory differently. Some distinguish near and far, some distinguish whether an item is near the speaker or listener, and some rely heavily on demonstratives, classifiers, articles or omitted arguments.
That linguistic variety matters because a target phrase cannot be chosen merely by replacing a source demonstrative with its nearest dictionary equivalent. If the source language distinguishes an object by distance but the target language distinguishes it by visibility or discourse status, the translator must preserve the intended identity using whatever construction is natural. A phrase may become more explicit or less explicit without becoming less faithful.
The visible scene supplies a reference map, not a license for creative guessing. Translators should reconstruct what the participants jointly knew. The intended referent may be a tool just picked up, a menu previously opened, or a region highlighted by a cursor. Correct identification connects grammar to visual evidence.
Make a small timeline of speech, gesture and action
In audiovisual work, the crucial evidence is sometimes spread across a few seconds. A speaker may say “use this” before lifting a tool, point at a screen after saying “there,” or name an object while the camera is changing angle. A translator working from a transcript alone can misassign the reference because the relevant visual event falls outside the subtitle segment.
Create a concise timeline: word spoken; gesture begins; cursor or hand reaches the target; action occurs; scene changes. This does not have to be a sophisticated timecode database. For a short clip, a note such as “12:04 ‘this’—12:05 cursor selects blue icon” can prevent a serious mistake. If the source and translated voiceovers differ in length, the target reference still needs to coincide with the same visual event.
Distinguish evidence from inference. A cursor hovering near an icon is weaker than a visible click on that icon; a hand moving toward a shelf is weaker than a hand grasping an object. If a shot cuts away at the decisive moment, say so in your working notes. Accurate translation begins by respecting what the footage actually shows.
This, that, these and those can encode different distinctions
English demonstrates a familiar near-versus-far contrast, but even in English the contrast is not purely physical distance. “This problem” may mean the issue currently being discussed. “That idea” may refer to something just mentioned. “These steps” can indicate a numbered procedure rather than objects at arm’s length. A camera demonstrator may say “that” while holding the object close because the object was introduced earlier in discourse.
Other languages may include distinctions involving the listener, accessibility, visibility or social relationship. A mechanically chosen target demonstrative can make an item sound like it is located elsewhere or already known when it is not. Begin with the referent, then select the target construction based on the target language’s own system. The number of words may change.
If the spoken source is vague and two objects are visible, do not solve the ambiguity by choosing the grammatically most specific target pronoun. Translate the uncertainty honestly and, if the audience must act, see whether the surrounding shot establishes one intended item. Functional clarity should come from evidence, not editorial confidence.
Pronouns can point to events and not just objects
Consider “That was the difficult part.” The pronoun that may refer to tightening a screw, making a decision, the preceding explanation or the entire previous stage of a demonstration. It is not necessarily a visible physical item. A target language that asks for an explicit noun may require a choice such as “that step” rather than “that object.” The preceding clause and action sequence determine the answer.
Similarly, “Do it again” can refer to the immediately completed movement, the whole procedure or a particular operation demonstrated earlier. Translate the action boundary carefully. A line that tells the viewer to repeat the entire process when the speaker meant one adjustment can waste materials, damage a product or confuse a learner.
When unsure, summarize the preceding event in your working notes before choosing the target wording. This is a practical test of comprehension: if you cannot say what “it” refers to in the source, you cannot know whether the target pronoun or noun will preserve the same action. Reference recovery is a prerequisite for fluent translation.
Camera position can reverse everyday spatial language
“Left” and “right” seem objective until you ask whose viewpoint controls them. The speaker may mean their own left, the viewer’s left, the object’s left side, or the left side of an image as displayed. A mirrored front-facing camera can invert what the viewer sees. A video editor may horizontally flip footage after recording, leaving the words and image inconsistent.
Do not assume the translator should silently correct such contradictions. First determine whether the source intentionally refers to the performer’s viewpoint or the viewer’s. For a task-oriented target audience, a clarifying phrase such as “on your right” may be justified if the reference is confirmed. If the original is demonstrably wrong, flag the issue to the publisher and preserve editorial accountability rather than disguise it.
For procedures, pair spatial words with stable anchors when possible. “The switch next to the red indicator” can be more reliable than “the switch on the left” after different camera crops. But the anchor must be visible and accurate. The translator should not add a color, label or object relationship that is absent from the actual frame.
Above, below, behind and next to are relational
Spatial prepositions depend on coordinate systems and object orientation. “Above the line” in a graph, “above the button” on a screen and “above the bowl” in a kitchen use different physical relationships. A word-for-word equivalent may be unnatural or misleading in the target language because of its usual spatial conventions. The translator must identify the actual relationship before selecting the expression.
Software demonstrations are particularly sensitive. The speaker may say “the menu below this tab” while a responsive interface later repositions the menu. A screenshot, not just a transcript, is needed to know whether the instruction is still actionable. If the localized interface uses a different layout, the target script may need an approved adaptation rather than a literal spatial translation.
In educational material, relational language can be the point of the lesson. A geometry instructor contrasting “inside,” “outside,” “adjacent” and “opposite” is teaching precise conceptual relationships. There the translator should prioritize stable technical meaning over a casual demonstrative shortcut. A viewer must be able to reconstruct the same spatial relation from the target words.
The index finger is not a universal label
A point can indicate an object, a route, a region, a movement direction or someone who should speak next. Its social meaning also varies across communities. However, in this task the central issue is not the cultural interpretation of pointing as a gesture; it is the way the visible gesture resolves a linguistic reference. The difference matters. “Put that here” may contain two separate gestures with two separate anchors.
Describe the physical evidence first: the hand extends toward the blue lid; a second gesture indicates the shelf. Then infer the sentence’s mapping: “that” means the lid, “here” means the shelf location. Translate the instruction naturally while preserving both roles. If the hand is out of frame, do not assume the reference from the direction of the forearm alone.
When gesture interpretation is itself the topic—whether a beckoning sign is polite, for example—use a different translation analysis. eduKateSG’s broader materials on cultural meaning and context address that question. Here, the gesture primarily answers “which thing?” or “which place?” so that a practical instruction remains attached to its object.
Cursors, highlights and animation are visual pointing
In a screen recording, a cursor can play the role of a finger. So can a glowing border, a mouseover state, a pulsing dot or an animated arrow. The narrator may say “choose this” exactly as the highlight moves. The relevant visual anchor is the interface state at that moment, not a stable object name printed elsewhere on the screen.
A target-language subtitle can become longer and outlast the highlight. This creates a new referential error: viewers read “this one” after the indicated item has disappeared. A localization team may need to retime the animation, shorten the target sentence without losing the instruction, or position the subtitle so it remains interpretable. Translating the sentence in isolation will not solve the synchronization problem.
When localizing the software itself, also check whether the relevant button label changes in the target UI. A subtitle should match the visible version the learner actually uses. For detailed interface translation, see eduKateSG’s website, app and interface translation guide. Screen references need both correct language and correct on-screen affordances.
Object identity can survive a cut even when position changes
A camera cut can move the same object from left to right, enlarge it, or remove surrounding landmarks. If a speaker says “keep that there” across the edit, the intended object may be the one handled before the cut rather than the object now nearest the center of the screen. A translator should follow object continuity through the action, not rely on a static coordinate.
Look for distinctive features: color, shape, attached label, relation to a hand, and the role it plays in the task. If two identical cups exchange places, visual continuity may be genuinely uncertain. Mark that in production notes. The target translation should not invent a distinction such as “the taller cup” when no such feature is visible.
This problem is common in demonstrations assembled from multiple takes. The source audio may have been recorded after filming or revised during editing. Treat the final published video as the operative object the audience encounters, while using original project materials to resolve discrepancies when available. An accurate target must align with the footage people will actually watch.
A missing object requires restraint, not invention
Sometimes the speaker points outside the frame. “Put it over there” refers to a location the audience cannot see. A translator may be tempted to convert the phrase into “put it on the shelf” because a shelf appears in another shot. But unless the source context establishes that connection, the shelf is an unsupported addition. Better to retain “over there” or ask the producer for clarification.
In accessibility-focused descriptions, it can be useful to identify an off-screen referent when the video or accompanying materials genuinely support it. But an explanatory track should distinguish description of visible evidence from a reconstruction inferred from a later action. The line “the presenter points off-screen” is not equivalent to a confirmed instruction to use a particular shelf.
If the ambiguity prevents safe operation, do not force the translation to look complete. Mark the original communication problem. The source video may need a reshoot, a close-up or a graphic label. Language alone cannot manufacture missing visual evidence without creating the risk of directing users to the wrong place.
Case study: choosing the correct icon in a software tutorial
A tutorial displays a toolbar with a square icon and a circular icon. The presenter says, “Click this one to switch modes,” while the cursor rests over the circle. A literal translation of “this one” may be perfectly natural, provided the viewer can still see the cursor and understand the instruction. Expanding it to “click the square” would be an inaccurate inference.
Now imagine the localized interface reorders the toolbar. The same target-language subtitle could still be linguistically correct while functionally wrong if its timing or associated highlight no longer identifies the circle. The team needs to review the actual target screen. UI labels, icon positions and voiceover timings should be tested together rather than reviewed by separate people without a shared reference.
When the tool has an accessible name, consider whether saying “the Circle Mode button” improves usability. The label should be verified from the software, not invented from visual appearance. The best translation preserves the intended action and gives the user an unambiguous way to perform it.
Case study: a cooking video with two bowls
A cook places flour in one bowl and sugar in another, then says, “Pour this into that.” A camera close-up shows the cook lifting the flour bowl toward the sugar bowl. The pronouns are semantically thin, but the movement identifies both source and destination. A target language may prefer nouns to avoid an unclear pair of demonstratives, yielding “pour the flour into the sugar bowl” if the contents are clearly established.
Do not add the ingredient identities if the video never verifies them. Two white powders can look similar on camera, and a translator cannot infer composition from appearance alone. A recipe card or earlier narration can provide stronger evidence. If ingredients are dangerous to confuse, confirmation from the source recipe is essential.
Notice also the direction of action. “This into that” has two argument roles. Reversing them changes the procedure. A translator should mentally draw an arrow from the lifted container to the receiving one and compare that arrow with the target sentence. This simple check is often more valuable than debating the stylistic elegance of pronoun choices.
Case study: assembly instructions and mirrored views
An instructor holds a bracket toward the camera and says, “The notch goes on this side.” Their own right hand touches one edge, but the camera mirrors the image. If subtitles specify “right side,” a viewer may attach the bracket backward. The safe solution begins with verifying the physical feature: which edge has the notch, where does it face relative to the assembled object, and whether the recording has been mirrored.
A source transcript might not mention right or left at all. That does not entitle the translator to add one from a misleading visual orientation. A more stable target expression can identify the edge by a visible feature or by the attachment point, provided the documentation confirms that relationship. If no stable anchor is available, flag the footage for correction.
This case shows why target-language naturalness is subordinate to procedural truth. A person following the instructions needs to create the same configuration as the instructor intended, not merely understand the grammatical relation between “this” and “side.”
Case study: a mathematics lesson pointing at a graph
A teacher says, “This increases while that stays constant,” pointing to two lines on a graph. The lines may cross, change colors or be represented by a legend elsewhere. A translation that treats “this” as the horizontal axis and “that” as the vertical axis because of the teacher’s hand position could reverse the mathematical relationship. The key question is which plotted quantity the teacher intends.
Follow the pointer at the exact time and inspect the prior explanation. If the teacher has named two variables, the target language may need to repeat those variables to make the contrast accessible. Such explicitness can be faithful because it preserves the identification already available in the classroom context. It should not introduce a new scientific claim.
For students, this is a useful reminder that diagrams are part of language comprehension. Vocabulary and grammar help interpret the sentence, but the graph establishes the referent. The target lesson should allow a learner to identify the same rising quantity and the same constant quantity without relying on the translator’s guess.
Case study: a craft demonstration with hidden hand movement
In a sewing demonstration, the camera frames only the top of a fabric piece. The presenter says “fold this under,” but the crucial movement happens beneath the fabric. A translator might infer that the hem, seam allowance or entire strip is being folded. Those are different operations. If the video later reveals the final construction, it may clarify the step; if not, there is no reliable basis for selecting a technical noun.
The correct response depends on publication purpose. An ordinary subtitle can preserve “fold this under” if the source itself uses that phrase. An instructional adaptation intended for people who cannot see the movement needs more than translation; the producer may need a clearer camera angle or an additional verified description. Do not hide the missing demonstration behind an apparently precise target phrase.
A good quality review asks whether a beginner who reads the target instructions can reproduce the intended movement. If success depends on visual evidence absent from the finished video, the translation process should report that problem instead of treating it as a minor wording issue.
Case study: medical or laboratory procedures
High-stakes demonstrations require stricter referential verification. “Insert this here” is not safe guidance if several ports, vials or controls look similar. Translators should consult the approved procedure, equipment labels and qualified subject-matter reviewers rather than choose a referent from a quick visual impression. A wrongly named connection point can cause more harm than an awkward sentence.
Preserve units, sequencing, warnings and conditions exactly. If the source says “only after the indicator changes,” the target should not shorten this into an unconditional action to save subtitle space. A visual demonstration may help locate the object, but it does not replace the technical instruction. When evidence conflicts, the discrepancy must be resolved with the authoritative source before release.
This article explains a language method, not a substitute for professional procedural review. The principle generalizes: as the cost of acting on the wrong referent rises, independent verification becomes more important. The translator’s job includes knowing when a missing visual anchor requires escalation.
Case study: live presentations with a laser pointer
A presenter says, “Notice this peak,” while a laser dot flashes near a chart and disappears. The translation may be displayed several seconds later by live captions, after the pointer has moved. A viewer who depends on captions can no longer identify the peak. The error is temporal rather than grammatical: “this peak” could be an accurate phrase at the wrong moment.
For recorded material, align the target caption with the relevant highlight. In live settings, a captioner or interpreter may need the speaker to name the axis value or category verbally. Presenters can improve multilingual access by replacing unexplained pointer-only statements with phrases such as “the second peak near the middle of the graph.” The source speaker should provide that clarification rather than expecting translation to guess.
This is a useful design lesson for presenters everywhere: precise verbal anchors benefit listeners, people with visual disabilities and multilingual audiences simultaneously. The more the original instruction depends on a fleeting beam of light, the harder it is to communicate robustly through delayed text.
Case study: “here” in navigation and wayfinding
A guide in a recorded walk says “turn here,” and the shot shows a junction from the guide’s moving point of view. If the audience watches later or approaches from another direction, “here” depends on the path being demonstrated rather than on the viewer’s current location. Translating it as an instruction directed to an arbitrary reader can create confusion.
Identify the reference frame: route-following from the guide’s viewpoint, location on a map, or actual viewer position. When the video explicitly demonstrates a route, natural target wording may include “at this junction” while retaining the visual context. If a named landmark is shown and verified, it can be used as a stable anchor. Do not invent a street name or assume the route remains unchanged over time.
For live services, map interfaces may localize orientation terms differently from ordinary conversation. A rightward gesture, compass direction and screen-right arrow are not interchangeable. A correct translation should maintain which action the original navigation actually instructed at that point in the route.
When a speaker points to text instead of a physical object
A teacher may underline a sentence and say, “This word tells us the tense.” The referent is a written token in the image, not merely its dictionary meaning. If the target edition translates the underlined sentence, the corresponding word may shift position or become a multiword expression. The voiceover and highlight must be rebuilt around the target-language grammar rather than copying source coordinates.
Sometimes the teacher’s lesson is about the source language itself. Then the original word should remain visible and may need a gloss alongside it. Replacing the word with a translated equivalent could eliminate the feature being taught. Distinguish a language-learning demonstration from ordinary content delivery before changing any on-screen text.
The same issue appears in code tutorials, proofreading videos and document reviews. “Remove this comma” refers to a specific punctuation mark; replacing it with a target punctuation convention without adapting the screen image makes the instruction incoherent. Translation should track the instructional object, whether the object is a tool, word, symbol or region of the screen.
Subtitles, captions and voiceover need different solutions
Subtitles are short written overlays that share the screen with the visual scene. They should use concise natural language, but compression must not change which object is indicated. Captions may additionally represent speaker identity and relevant non-speech sounds. Voiceover is spoken language and may require retiming to coincide with an action. The same literal translation can therefore fail differently in each medium.
For subtitles, check readability at the actual display size and scene duration. For voiceover, read the translated line aloud against the action. For a transcript, add verified references in a separate explanatory note when necessary because the still text lacks the accompanying frame. Each medium preserves the same underlying referent while solving a different presentation constraint.
Avoid duplicating every visible gesture in text. A subtitle does not need to narrate “the speaker points” when the pointing is obvious and the speech is already clear. But if the gesture carries the only information identifying the target, the deliverable may need appropriate clarification or an accessibility description. Choose the least intrusive method that preserves usability.
Translation memory and segmentation can detach the pointer
Translation tools often divide content into segments. A segment containing “this one” may appear without its preceding explanation, frame or slide identifier. A reused translation can then attach the phrase to a different object in a later video. Translation memory is valuable for stable terminology, but deictic phrases are highly context dependent and poor candidates for blind reuse.
Store screenshot references, timestamps, segment IDs and the speaker’s action alongside the source text when feasible. Review repeated phrases individually even when a tool reports a perfect text match. Two identical English strings can have different target translations because one points to a singular item and another points to a group or earlier event.
Workflow design can reduce errors. Give translators access to time-synced video rather than isolated spreadsheet rows; let reviewers flag unresolved referents; and preserve the link between the target subtitle and the frame where it is supposed to make sense. This is not excessive process for a complicated sentence. It is the minimal context necessary to know what the sentence means.
A systematic visual-reference quality check
Review every target demonstrative, spatial term and pronoun that instructs an action. Ask whether its referent is visible, whether it is the same one as in the source, whether the speaker’s viewpoint is preserved, and whether the action timing remains synchronized. For screen content, check the final localized interface rather than the original mock-up. For edited video, inspect both sides of every relevant cut.
Make a second pass with subtitles hidden and then visible. The first pass helps you understand the actual action without reading potentially misleading text. The second reveals whether the translation competes with or supports that action. A reviewer unfamiliar with the source language can sometimes detect a reference mismatch by following the target instructions alone.
When the source itself is ambiguous, label the ambiguity instead of disguising it. A translation can be accurate while exposing a source-production problem. Quality assurance is the point at which that problem should become visible, with enough detail for the creator to correct it before viewers are asked to act.
How students can practise visual reference
Use short, ordinary clips in which a speaker points to two or three safe objects. First mute the clip and describe the visible action. Then listen to the speech and identify every referring expression. Finally write a target-language sentence and ask a partner to choose the intended object using only the target line plus the picture. If the partner consistently selects the wrong object, inspect the reference choice.
Next remove the video and work from a transcript. Learners will notice which meanings can no longer be determined. That exercise teaches epistemic restraint: it is sometimes impossible to translate a deictic instruction confidently without the scene. Try the reverse as well. Play the video but cover the spoken subtitles to observe how much context is carried visually.
This work connects with receptive vocabulary learning. Understanding individual words helps, but interpreting a full message requires binding those words to objects, actions and prior discourse. A learner becomes more accurate by asking what each expression refers to rather than searching for isolated equivalents.
Frequently asked questions
Can I translate “this one” without seeing the video? Sometimes, if the surrounding transcript identifies the referent unambiguously. If not, request the relevant frame or preserve uncertainty rather than inventing an object.
Should subtitles replace vague words with object names? Only when the names are supported by the video or another authoritative source, and when the additional specificity helps rather than changes the source. Some languages use demonstratives naturally where others prefer repeated nouns.
What if the video was mirrored? Verify the real viewpoint and the final released footage. Do not silently reverse left and right from intuition; correct or clarify the instruction with the producer when necessary.
Does a pointing gesture have the same meaning everywhere? No. But this article’s focus is narrower: how the visible point identifies the object of a spoken phrase. Cultural interpretation of the gesture is a related but separate question.
What if the source instruction is wrong? Flag the mismatch and follow an agreed editorial correction process. A translator should not hide a source error by quietly making up different content.
Further analysis: demonstratives in a multilingual exchange
Imagine two people around a table. One stands beside the window and the other beside the door. The first says “that” while the second reaches for an object near the door. Some languages distinguish location relative to speaker, listener or both. A target sentence that maps the source word to a fixed near/far pair may fail even though both versions are grammatically acceptable.
The translator should diagram the participants’ positions and ask which item the listener was expected to understand. If the target language has a listener-proximal demonstrative, that form may express the same intended location more precisely than a word chosen solely for phonetic resemblance. If the target language relies on context, a plain noun phrase may be more natural. The goal is referent identity, not preservation of source grammatical packaging.
This is also why viewer position differs from participant position. The camera places an audience at a third vantage point. A subtitle must remain intelligible to that audience while not falsely rewriting the relationship between the participants. The translator can sometimes use a simple descriptive phrase to bridge the different viewpoints.
Further analysis: preserving ambiguity when the scene is deliberately mysterious
In mystery stories, suspense often depends on not revealing what “that” refers to. A character may point off screen and say, “Did you see that?” while the camera shows another character’s reaction rather than the object. An over-explicit translation such as “Did you see the person behind the tree?” could spoil the scene and invent a fact the audience has not been given.
Educational demonstrations generally reward clarity; narrative scenes may intentionally withhold information. Therefore the translation brief matters. Ask whether the audience is meant to perform an action, learn a concept or experience a dramatic uncertainty. In the last case, ambiguity is part of the source meaning and should survive naturally in the target language.
A skilled translator does not maximize explicitness in every sentence. They preserve the amount of information the source makes available at that moment. The reference may be resolved later; let the target viewer discover it at the same stage. This is fidelity at the level of audience experience, not merely noun choice.
Further analysis: object tracking across a sequence of instructions
A long demonstration often reuses pronouns over several steps: “Put it here. Turn it around. Then take that and attach it.” The objects may move, and “it” may change referents when a new tool is introduced. A translator should build a minimal object ledger: item A placed on base B; item C rotated; component D attached to A. This prevents an earlier noun from taking over a later pronoun by inertia.
The object ledger is particularly valuable when the target language requires grammatical gender, noun classes or classifiers. The translator may have to specify a class marker that forces a referent choice not explicit in the source. That choice should follow the action sequence, not the most recent noun in the transcript. Sometimes the right solution is to repeat a noun rather than produce a misleading pronoun.
After drafting, walk through the sequence as a first-time viewer. At every step ask which object is moved, where it goes and whether that is still the object shown in the video. An instruction chain is only as reliable as its weakest reference handoff.
Further analysis: accessibility without overloading the scene
Some audiences cannot easily see a cursor, pointing finger or small text highlight. The source “choose this one” may therefore be inaccessible even when a sighted viewer understands it immediately. Accessible adaptation can add a verified name, position or label so the instruction does not depend exclusively on the gesture. But the addition belongs to an explicit accessibility purpose, not an unacknowledged translator guess.
Good description is selective. It identifies the necessary object and action without reciting every irrelevant movement. If a speaker points to the green Submit button, “select Submit” may be sufficient when the button’s accessible name is Submit. Adding the exact shade of green may not help. If two Submit controls exist, include another verified distinguishing feature.
Accessibility review should use the same evidence standards as translation review: what is visible, what does the software expose to assistive technology, and what action should the user take? The strongest result makes a task independently understandable across languages and access modes while remaining faithful to the source.
The guiding idea
A word such as “this” is not a complete reference until the scene supplies its target. Translate it only after you know what it points to. Track the relevant gesture, cursor, object, participant viewpoint and action timing; choose a natural target construction; and test whether a viewer can identify the same item and perform the same step.
The surrounding skills belong to a larger learning route through context, discourse, translation units and segmentation, and the How to Translate Easily series. The practical criterion remains concrete: the translated words should lead the viewer’s attention to the same thing at the same moment, without requiring an invented visual fact.