How to Understand English on Video Calls

Struggling to understand English on video calls? Diagnose audio, latency, overlap and caption problems, then repair only what you missed and verify key details.

Scarlett Vale

FunFluen editor · Film and listening

Aug 24, 2026 22 min

The short answer

To understand English on video calls, first decide whether the failure is language, audio, speaker tracking, or latency; wait a beat, identify the speaker, ask for only the missing unit, and put names, numbers, decisions, and actions in writing.

A video call goes quiet. You start answering. A moment later, someone else’s voice arrives and you both stop. It is easy to think, “My English is too slow.” Sometimes language is the problem — but the failure may also be audio, speaker tracking, latency, or overlap. Don’t grade your English until you diagnose the channel. Use See → Wait → Name → Verify → Write to recover the missing piece without asking the whole meeting to rewind.

What failed?

Language: the audio is clear, but a word, phrase, or meaning is unknown. Ask for the meaning or a short rephrase.
Audio: sound clips, freezes, drops, or disappears. Ask only for the lost sentence, name, or number.
Speaker: the message makes sense, but you cannot attach it to the right person. Identify or verify who said it.
Timing: silence suddenly becomes two voices at once. Leave a beat, yield explicitly, and restore one-speaker-at-a-time talk.

See → Wait → Name → Verify → Write. See who is speaking. Wait a deliberate beat. Name the speaker or missing unit. Verify the important detail. Write down anything that cannot afford to be wrong.

What a video call gives you and takes away

A video meeting gives you more than sound: name labels, an active-speaker tile, optional camera cues, captions, and a main chat. Great. It also gives you frozen video, late audio, accidental overlap, hidden mute states, and captions that can look perfectly readable while being wrong. The trick is to use the extra channels without trusting any single one too much.

Start with See. Attach each point to a person using the active tile, a name label, the voice, or an explicit verbal handoff. Then keep a tiny live map — not formal minutes, just enough to stay oriented:

If your problem is a call with no visual channel at all, use the separate guide to understanding English on the phone. If you lose track of speakers and topic changes even in ordinary face-to-face groups, the better fit is the guide to tracking multi-speaker conversations. Bad conference-room acoustics are another problem again; this page stays with multi-party video calls.

Diagnose the failure before choosing the repair
Failure	Visible clue	What not to assume	Smallest repair
Language gap	The sound is clear, but one expression or the point itself is unfamiliar.	Do not assume the microphone or connection failed.	Ask for the unclear meaning or one-sentence rephrase.
Audio break-up	A voice clips, crackles, freezes, drops syllables, or disappears.	Do not assume your vocabulary suddenly failed in the exact same instant.	Request only the lost sentence, name, or number.
Wrong speaker	You understand the idea but cannot tell who said it or who owns it.	Do not attach the point to the person whose tile happens to look active.	Name the person you think spoke and verify the attribution.
Latency gap misread as silence	A pause feels complete, then another response arrives just as you begin.	Do not assume the other person ignored you, had nothing to say, or that you were too slow.	Leave a deliberate beat before taking the turn.
Overlapping turns	Two or more voices become audible together.	Do not try to decode both streams at once.	Yield explicitly or ask for one speaker at a time.
Hidden mute	A person appears to be speaking, but no voice arrives; others may react to the mute state.	Do not assume your headphones or listening ability failed first.	Check the visible meeting state and wait for the speaker to restore audio.
Caption error	The caption is plausible but conflicts with the sound, context, spelling, or later discussion.	Do not treat readable automatic text as verified truth.	Verify the consequential unit through read-back, chat, or follow-up.
Missing owner or deadline	You understand the decision but not who will act or when.	Do not assume the action belongs to the last person who spoke.	Verify owner plus deadline and move them into writing.

The audio is clear, but one phrase makes no sense. What failed?

Most likely: language or meaning. The channel delivered the sound, so asking everyone to troubleshoot microphones wastes time. Ask for the unclear point to be rephrased or identify the exact expression that blocks the meaning.

You hear “The revised total is forty—” and the voice cuts out. What failed?

Audio. You already have the context and part of the sentence. Do not request the whole explanation again. Ask only for the missing number, then put that number in the main chat if it matters.

You understand a proposal but cannot tell who made it. Is that a vocabulary problem?

No. It is a speaker-tracking problem. Your semantic understanding may be fine while attribution is missing. Verify the speaker or action owner before adding the point to your live map.

A colleague’s mouth is moving but there is no sound. Should you assume your listening failed?

No. First check the meeting state and the group’s reactions. A hidden or accidental mute is a channel problem. A beautifully prepared sentence can occasionally become silent theatre; your English did not cause that.

That diagnosis matters because one of the least obvious failures is not missing sound at all. It is receiving a voice late.

Latency destroys turn-taking

Conversation normally relies on tiny timing cues: a voice trails off, a pause opens, and someone else comes in. Video transmission can bend those cues. The important point is not that every pause is “lag.” It is that delay can make a normal remote response arrive after you have already interpreted the silence.

The official summary of ITU-T Recommendation P.1305 on telemeeting delay says delay can affect interaction without participants explicitly noticing it, and people may attribute the effect to the other participant’s behaviour. The effect is task-dependent, so there is no honest universal number at which every call suddenly becomes “too delayed.”

A 2021 study by Seuren and colleagues analysed 25 UK video consultations in heart failure, antenatal diabetes, and cancer services, totaling 7 hours 57 minutes. In those consultations, latency could create perceived silence, overlapping talk, and difficulty restoring one-speaker-at-a-time conversation. That healthcare setting is important: it supports the mechanism, not an exact prediction for every workplace meeting. Read the Seuren et al. study.

Boland and colleagues also found that small, variable transmission delays disrupted normal conversational timing and delayed turn initiation in controlled and unscripted comparisons using Zoom. That does not mean Zoom is uniquely bad or that every online interruption is caused by the platform. It means timing itself can become less trustworthy. See the Boland et al. paper.

This is the Wait step: after a turn seems finished, leave a deliberate beat before you claim the floor. You are not waiting forever. You are giving a travelling response a chance to arrive.

The call goes silent after a question. Should you jump in?

Usually, give it a deliberate beat first. The silence could be ordinary thinking time, a delayed response, a mute problem, or genuine completion. Waiting briefly costs little and can prevent the classic collision where two people decide the silence belongs to them at the same moment.

A colleague answers late. Did they ignore your question?

You cannot conclude that from timing alone. Delay can alter interaction without being obvious, and a person may also be thinking, checking information, or dealing with audio. Treat the timing as a clue, not a character judgment.

A beat prevents some collisions. Not all of them. When two voices still arrive together, the goal changes from “understand everything” to “restore one clean stream.”

Overlap and who yields

The world’s most polite video-call traffic jam sounds something like this: one person yields, the other yields, both restart, both stop. The answer is not heroic simultaneous listening. It is a small turn repair.

Use Name here: attach the point to a person, and name the exact unit you missed. The nine phrases below are deliberately bounded listening repairs — not a full business-English phrase list.

Nine minimal listening repairs for three common breakdowns
Breakdown	Phrase	Use it when…
Audio / mute / break-up	“You’re breaking up. Could you repeat the last sentence?”	The signal damaged the most recent sentence and you need that sentence again, not the entire discussion.
	“I caught [known part], but I missed the name/number after it.”	You understood most of the turn and can identify the precise missing unit.
	“Could you put that name or number in the main chat?”	The detail is easy to mishear or important enough to deserve a checkable written form.
Latency / overlap	“I think we spoke at the same time — please go ahead.”	You collided with another speaker and yielding is the fastest way to restore one stream.
	“I’ll pause here. [Name], were you coming in?”	You suspect a named person began speaking during the timing gap and want to hand the floor back clearly.
	“Could we take this one speaker at a time?”	Overlap continues and the group needs a shared reset. This is more directive, so hierarchy, meeting culture, and your role can affect how readily you use it.
Meaning / action still unclear after audible audio	“Just to check, the decision is [X], right?”	You heard the discussion but need to verify the final decision rather than replay the whole argument.
	“Who owns that action, and when is it due?”	The task is clear but responsibility or deadline is missing.
	“Could you rephrase the last point in one sentence?”	The audio is audible, but the meaning remains unclear and a shorter restatement would help.

If you are junior in the meeting, speaking across a sensitive hierarchy, or working in a culture with stronger turn-taking norms, yielding first may be the lower-friction choice. That does not make the other phrases “wrong”; register and status affect how direct a repair should feel.

Two tiny English traps worth fixing

Repair wording: what the listener hears versus what you probably mean
Original expression	Classification	What a listener understands	Likely intention	Natural alternative	When the original still works
“Sorry?”	Context-dependent, not wrong	You need some kind of repetition or repair, but the scope is unclear.	You want the part you missed repeated.	Use the relevant precise repair from the matrix, such as naming the missed name, number, sentence, decision, owner, or deadline.	It works well when the entire short turn was missed or the missing scope is already obvious from context.
“I’m in mute.”	Unusual / non-idiomatic in ordinary meeting English	The listener will probably infer that your microphone is muted.	You mean that your microphone is currently muted.	“I’m on mute.”	There is no ordinary workplace-English context where “I’m in mute” is the standard collocation; if it appears, treat it as nonstandard wording or a technical string rather than normal conversational English.

The voice breaks only during a surname. Which repair category fits?

Audio / break-up. You already understand the surrounding meaning. Ask only for the surname, and if spelling matters, move it into the main chat rather than requesting the whole explanation again.

You and a senior colleague start together. Must you always insist on one-speaker-at-a-time talk?

No. The repair depends on the moment and your role. A quick explicit yield is often enough after one collision. Asking the group to take one speaker at a time is useful when overlap persists, but hierarchy and local meeting norms still matter.

Production challenge: you heard the sentence but missed only the client’s name. Is “Sorry?” the best repair?

It is valid but too vague for this situation. “Sorry?” tells the listener that something failed, but not what. A better move is the matrix pattern that says what you caught and names the missing unit. That protects the meeting’s momentum because the speaker only has to repair the client name.

For a broader lesson on what to say throughout calls — openings, closings, requests, and general phrase production — use the separate guide to English phrases for phone and video calls. Here, the language stays narrow because the listening job is narrow.

Once the voices are clean again, do not make audio carry information that cannot afford to be wrong. That is where Verify → Write takes over.

Use chat as a second channel

The main meeting chat is not just for “hello everyone” and the mysterious link someone promised twelve minutes ago. It can be your second listening channel.

Use it for fragile details: names, numbers, links, decisions, owners, and deadlines. Those are the pieces that matter when you heard almost the entire sentence but the missing piece was the important one.

Main-chat drill: move six fragile details into writing

These are model messages for an illustrative meeting. Notice that they are short enough to support listening rather than turning you into the meeting secretary.

Model main-chat messages
Detail	Model message	Why it helps
Name	“Name check: Amina Rahimi — is that the correct spelling?”	A name can be consequential when spelling or identity matters, so a written check prevents you from relying on one audio or caption pass.
Number	“Confirming the total: 14,500.”	The group can correct a digit immediately.
Link	“Could you paste the link in the main chat?”	You do not need to decode or remember a spoken URL.
Decision	“Decision check: the launch moves to Thursday.”	A visible summary separates the final decision from earlier options.
Action owner	“Owner check: Marco will send the revised file.”	Responsibility no longer depends on remembering who spoke last.
Deadline	“Deadline check: 16:00 UTC on Thursday.”	Time, date, and time zone become inspectable instead of fleeting audio.

This is not a full minutes system. Keep your live map tiny. The purpose is to preserve the meeting’s meaning while you listen, not to multitask so aggressively that you miss the next three turns while documenting the previous one.

Should every sentence go into chat?

No. Write the details that are fragile or consequential: names, numbers, links, decisions, action owners, and deadlines. If the point is easy to hear, low-risk, and already clear, writing it adds cognitive load without much benefit.

You heard a contract deadline once and the captions agree. Is that enough?

Not for a consequential detail. For contracts, money, employment, safety, access, legal or medical matters, deadlines, and similar high-impact information, use read-back, main-chat confirmation, or written follow-up. Agreement between your ear and an automatic caption is useful evidence, but it is not the same as confirmation.

Writing solves persistence. It does not always solve speaker identity. A clear video feed can sometimes help you see who is taking the turn — but that is a request, not an entitlement.

Ask for the camera on

A camera can sometimes give you partial cues: who is speaking, who is about to come in, a gesture, or visible mouth movement. Do not turn that into a promise of lip-reading or guaranteed comprehension. Video is one cue among several.

If it would genuinely help you track the speaker, a bounded request is:

“If you’re comfortable and your connection allows, could you switch your camera on while you explain that part? It helps me track who is speaking.”

The first half matters as much as the second. Camera-off may be necessary or preferred because of bandwidth, privacy, disability, fatigue, safety, culture, organizational policy, or the person’s environment. Camera-off does not prove disengagement.

If the camera stays off, switch methods: use names, explicit handoffs, raise-hand controls where available, the active-speaker indicator, and main chat. If video makes the connection worse, prioritize intelligible audio and the written second channel. A gorgeous frozen face is less useful than a clear sentence.

The camera stays off after you ask. Should you ask again?

Usually, switch channels instead of escalating the request. You do not know whether the reason is bandwidth, privacy, disability, fatigue, safety, culture, policy, or something else. Use names, explicit handoffs, platform turn controls, the active-speaker indicator, and chat. If video itself destabilizes the connection, audio is the priority.

Visual cues can support orientation. Automatic text can support it too — but “the captions caught it” is not the same sentence as “the detail is correct.”

Recordings, captions, and transcripts

Platform notes in this section were checked 24 August 2026. Menus, languages, plans, licenses, organizer controls, and client behaviour can change, so use the durable principle first: captions and transcripts are support channels, not authorities.

Current caption and transcript distinctions relevant to listening
Platform support page	What matters for this listening method	Do not assume
Microsoft Teams	Microsoft documents real-time live captions, recommends setting the correct spoken language for better accuracy, says live captions are not saved, and treats transcription as a separate feature. Microsoft also documents human CART caption support.	Do not assume the same controls, storage, languages, or access exist for every account, client, or plan.
Google Meet captions	Google documents participant-local caption controls and says displayed caption history covers only the portion of the meeting while you were present with captions turned on.	Do not assume the caption history is a complete meeting record.
Google Meet transcripts	Google documents transcription as a separate feature with edition, host, language, and storage constraints; its help page says transcripts contain meeting speech, not chat messages.	Do not assume a transcript preserves the written second channel or that these rules apply to other platforms.

Automatic speech recognition can also be uneven. Koenecke and colleagues published a 2020 audit of five commercial ASR systems using 19.8 hours of US sociolinguistic interview speech from 42 white and 73 Black speakers; average word-error rates were substantially higher for Black speakers. That is important evidence that ASR errors can be uneven, but it is not a 2026 benchmark of Teams, Zoom, or Google Meet, and it cannot predict whether one current platform will fail for a particular speaker. Read the PNAS study.

Never treat an accent or dialect as an error, and never judge a colleague by caption quality. When automatic captions are wrong, that is not evidence that the speaker’s accent or dialect is wrong.

Captions caught it — but should you trust it?

The caption shows an unfamiliar proper name clearly. Trust it?

Use it as a clue, then verify. A clear-looking proper name can still be incorrect. If the identity or spelling matters, confirm it in the main chat or another trusted written source before using it in consequential work.

Your ear hears “fifteen,” but the caption shows “fifty.” Which one wins?

Neither wins automatically. A number is exactly the kind of fragile detail that deserves read-back or written confirmation. Do not resolve the conflict by simply choosing the channel you prefer.

The captions show who should do the task, but you missed the handoff. Is the owner confirmed?

No. An action owner can have employment, money, access, deadline, or accountability consequences. Verify the owner aloud and put the owner plus deadline into the main chat or written follow-up if the detail matters.

Recording and transcription create another boundary. Do not covertly record, transcribe, screenshot, or copy meeting chat because a tool makes it technically possible. Use the platform’s visible recording or transcription process, follow organizational policy, give any participant notice that is required, obtain any consent or permission required by policy or applicable law, and follow applicable law. The exact requirements vary, so this is not country-universal legal advice.

For accessibility-critical use, human CART captioning, interpreting, or organizational accessibility support may be appropriate where provided. Automatic captions can be useful, but they do not replace a required human accommodation.

Current support details: Microsoft: Use live captions in Microsoft Teams meetings; Google Meet: Use live captions; Google Meet: Use Transcripts.

The safest call is the one where you decide your fallback before the first awkward silence arrives.

Prepare the call

Your 60-second pre-call action

This is deliberately small. You are preparing to listen, not launching a second career in meeting administration.

Open the agenda and the main meeting chat.

Identify the names and roles most likely to speak.

Preload one or two likely terms or acronyms that could block meaning.

Test your speaker, headphones, and microphone where the platform allows.

Decide which details must go into writing if they appear: names, numbers, links, decisions, owners, deadlines.

Zoom support note, checked 24 August 2026: Zoom documents speaker and microphone testing before or during a meeting and provides a test-meeting path. Use that as setup help, not as a guarantee against connection problems. Zoom: Testing your audio settings for Zoom meetings.

Before, during, after

Before

Read the agenda.

Identify names and roles.

Preload likely terms and acronyms.

Test microphone, speaker, and headphones.

Join early enough to solve setup problems.

Open the main chat.

Decide what details must be written if mentioned.

During

Mark the active speaker in your live map.

Wait a deliberate beat at turn endings.

Yield explicitly after overlap.

Request the smallest missing unit rather than the whole meeting again.

Put names, numbers, links, and decisions into main chat when they need checking.

Verify the action owner plus deadline aloud.

After

Rebuild a short action record: decision, owner, deadline, unresolved point.

Compare it with the official summary or transcript if one exists and you are permitted to use it.

Request written confirmation for consequential ambiguity.

What failed? A five-beat call autopsy

The following meeting is fictional. Read each beat and decide what failed before opening the answer.

The audio is perfectly clear, but a colleague uses an unfamiliar expression and you cannot work out the intended point.

Reveal the diagnosis and repair

Language / meaning. The signal is intact. Ask for the point to be rephrased or identify the expression blocking meaning. Do not troubleshoot the network when the sound itself is clear.
A speaker says the project code; the middle of the code disappears in crackling audio, then the voice returns normally.

Reveal the diagnosis and repair

Audio break-up. You know exactly what kind of unit vanished. Ask only for the project code and put it in main chat if accuracy matters.
The facilitator asks for objections. There is a pause, you start speaking, and another response arrives at the same moment.

Reveal the diagnosis and repair

Possible latency / turn timing. The evidence is the silence followed by collision, not any weakness in vocabulary. Yield cleanly, then use a deliberate beat on the next turn. Do not claim latency with certainty; ordinary thinking time can produce a similar pause.
After the collision, both speakers continue for several words and neither stream is intelligible.

Reveal the diagnosis and repair

Overlap. Trying harder to listen will not turn two simultaneous streams into one clean signal. Restore one-speaker-at-a-time talk using the appropriate bounded overlap repair.
Automatic captions show a plausible client name, but the pronunciation you heard does not quite match it.

Reveal the diagnosis and repair

Caption / platform support uncertainty. The text is a clue, not confirmation. Verify the proper name in chat or another written source before using it. Never infer that the speaker’s accent or dialect is the problem simply because the recognizer produced uncertain text.

The pattern is the whole article in miniature: clear sound plus unknown meaning points to language; damaged sound points to audio; silence that becomes collision suggests a timing problem; simultaneous voices create overlap; plausible automatic text still needs verification when the detail matters.

Official audio lab: listen → classify → repair

Use BBC Learning English’s Office English: Calls and instant messages as the listening source, then use the official BBC transcript PDF for the checking pass. The 2024 material covers video-call problems such as mute states, slow internet, freezing video, breaking up, difficulty catching a line, polite interruption, and raise-hand controls.

Listen once without the transcript. When a problem appears, classify it as audio, overlap, or meaning. Do not pause to collect every phrase.
Listen again with the transcript. Find the smallest repair idea connected to the problem instead of copying the whole dialogue.
Choose one moment where your instinct would be a vague “Sorry?” and replace it with a precise missing-unit request from the nine-phrase matrix.
Open the model answers below and compare the diagnosis, not just the wording.

Model answer: mute or no audible voice

Classify it as audio / mute. The issue is not unknown vocabulary. The useful lesson is that meeting English contains technical-state language such as on mute, and the repair should stay with the missing audio rather than asking for unrelated explanation.

Model answer: slow internet, freezing, or breaking up

Classify it as audio / connection degradation. If the last sentence breaks up, the precise repair is: “You’re breaking up. Could you repeat the last sentence?” If the missing unit is a name or number, use the matrix pattern that identifies that exact unit and move it into chat if accuracy matters.

Model answer: people do not know when to speak and talk over one another

Classify it as overlap / turn-taking. The goal is not to listen harder to two voices at once. After a collision, one exact repair is: “I think we spoke at the same time — please go ahead.” A named handoff or the one-speaker-at-a-time repair can be used when the situation calls for it.

Model answer: you hear the words but still do not understand the point

Classify it as meaning. Audible audio does not guarantee comprehension. A precise repair is: “Could you rephrase the last point in one sentence?” If the meaning is clear but the decision or action is not, use the corresponding verification phrase from the matrix instead.

Model answer: turn a vague “Sorry?” into a precise repair

If you caught the context but missed a proper name or number, use: “I caught [known part], but I missed the name/number after it.” “Sorry?” is not wrong; it is simply underspecified when you already know which tiny piece disappeared.

Optional off-call practice with FunFluen

Once the live-call method is clear, you can practise the listening part away from a real meeting on suitable subtitle-bearing video. With FunFluen, the useful loop is small: listen to one checked line before reading it, replay it, move sentence by sentence, and slow a difficult line slightly before returning toward normal speed. This is deliberate practice support; it does not join meetings, capture live audio, read chat, fix networks, generate meeting captions or transcripts, identify speakers, or verify workplace decisions.

Review the FunFluen extension for video listening practice. Your first action is to review the extension listing before installing; this exact video-call lesson is not preloaded after the click.

Sources

The next awkward pause

The next time a call goes quiet and another voice lands on top of yours, do not turn that half-second into a verdict on your English. Ask a smaller question: What failed? If it was language, repair the meaning. If it was audio, recover the missing unit. If it was timing, wait and yield. If the fact matters, verify it and write it.

You do not need a perfect transcript of the meeting in your head. You need a reliable map of who said what, what changed, who owns the action, when it is due, and what still needs checking. For the broader skill beyond video calls, continue with listening at work.