Guides & Best Practices

Solving Speaker Desynchronization and Audio Lag in Hybrid Platforms: A Lip-Sync and Turn-Taking Guide (2026)

By the SmartMeet team · · 9 min read

Solving speaker desynchronization and audio lag in hybrid platforms comes down to two separate problems. The first is lip-sync: the voice and the face on screen arrive at different times, usually because video passes through more processing than audio. You fix it by measuring the offset with a clap test and delaying the faster signal on the sound desk until both line up. The second is turn-taking: people in the hall and people online talk over each other because each side hears the other a moment late. You fix that with headsets, a wired network and a clear speaking protocol run by the chair.

This guide is for event producers, AV technicians and meeting chairs running hybrid conferences, board meetings and summits in the UAE and beyond. It focuses on what happens between the stage and the online meeting. If your main concern is the delay that reaches remote interpreters and listeners, read why audio lag ruins remote simultaneous interpretation first; for echo and audio bleed, see how to fix echo and audio bleed in multilingual conferences.

Two problems that look like one

Delegates rarely say “the lip-sync is off”. They say “something feels wrong with the remote speaker” or “we keep interrupting each other”. Separate the two before you change any settings:

SymptomWhat is actually happeningWhere to fix it
Mouth moves before or after the wordsAudio and video are out of step (AV offset)Capture chain and sound desk delay
Remote speaker starts talking just as the hall doesEach side hears the other late (round-trip delay)Network, headsets and the chair's protocol
Remote speaker hears their own voice backEcho from the hall loudspeakers into the stage microphonesEcho and bleed control, not lip-sync
Interpreters lag far behind the speakerLong delay chain to the interpretersFeed path and interpreter set-up

Only the first two rows are covered in depth here. The other two have their own guides, linked above and in the symptom-by-symptom delay troubleshooting guide.

How much offset can an audience notice?

The usual reference is ITU-R Recommendation BT.1359, on the relative timing of sound and vision for broadcasting. Its subjective tests put the threshold of detectability at roughly 45 ms with sound ahead of the picture and 125 ms with sound behind it, and the threshold of acceptability at roughly 90 ms ahead and 185 ms behind. Two practical lessons follow:

  • Sound arriving early is worse than sound arriving late. We are used to hearing sound slightly after seeing its source, because sound travels more slowly than light. A voice that arrives before the lips move feels wrong very quickly.
  • Small offsets are fine; aim for “not noticeable”, not zero. You do not need laboratory precision. You need the offset inside the detectability window for the people watching the main screen.

These figures come from broadcast television tests, so treat them as a guide rather than a contract for a ballroom. A large hall adds its own delay too: sound from a loudspeaker at the back of a room takes time to reach the front rows.

Where lip-sync offset comes from in a hybrid set-up

In almost every hybrid venue the video path is slower than the audio path. Audio from a microphone goes through a digital mixer in a few milliseconds. Video goes through far more stages, and each one holds a frame or more:

Audio
  1. Stage microphone
  2. Sound desk
  3. Meeting computer
  4. Online meeting
Video
  1. Camera
  2. Video switcher or scaler
  3. Capture card
  4. Meeting computer
  5. Online meeting
  • Cameras and switchers. Processing, scaling and effects in a camera or production switcher add frames of delay. A chain of converters adds more.
  • Capture cards and software. The device that turns the HDMI or SDI signal into a webcam source buffers frames, and production software can add its own delay.
  • Projectors and LED walls. Image processing in the display itself delays the picture people see in the hall, so the in-room screen can look out of sync even when the online stream is fine.
  • Separate paths for sound and picture. If the meeting computer takes audio from the desk but video from a camera on a different device, the two take different routes and arrive at different times.
  • The network. Video and audio are carried separately over the internet and the receiving browser re-aligns them. A congested network makes this harder and can produce occasional drift for remote viewers.

Measure the offset with a clap test

You cannot fix an offset you have not measured. The simplest method needs no special equipment:

  1. Stand in front of the stage camera, next to a live microphone, and clap once, sharply, with your hands in shot. A film slate or a short, distinct tap on the lectern works too.
  2. Record what the online meeting shows on a second device that joins as a normal participant.
  3. Step through the recording frame by frame and note the frame where the hands meet and the point where the clap sound starts.
  4. The gap between them is your offset. At 30 frames per second, each frame is about 33 ms; at 25 frames per second, 40 ms.
  5. Repeat for each camera and for the in-room screen, because each path can differ.

Do this at the technical rehearsal with the real equipment, the real cable runs and the real internet connection. A test on a desk in the office will not show the delay of the venue's switcher or LED wall.

Fix lip-sync at the source

  1. Delay the audio to match the video. Most digital sound desks and many audio interfaces have an output delay. Set the delay on the feed to the meeting computer to the offset you measured, then repeat the clap test. Do not delay the hall loudspeakers to match the online stream: the hall should hear the live speaker naturally.
  2. Use one capture path. Where possible, send the meeting computer a single signal that already carries matched sound and picture, for example embedded audio on the camera or switcher output, rather than two separate sources.
  3. Shorten the video chain. Remove converters you do not need, and turn off effects and frame-rate conversions you are not using. Fewer stages mean less delay to compensate for.
  4. Wire the meeting computer. A wired connection to a dedicated line, not the attendees' Wi-Fi, keeps the stream steady so the receiving browser has less to correct.
  5. Check the in-room screen separately. If the LED wall or projector shows the remote speaker late, check whether the display has a low-latency or game mode, and match the hall audio of the remote speaker to it if needed.

Our walkthrough on connecting the venue audio to the online meeting covers the cabling for the floor feed and the return feed in more detail.

Stop the hall and remote speakers talking over each other

Even with perfect lip-sync, a remote panellist hears the hall slightly later than the people in the room do, and the hall hears the reply slightly later again. In a fast debate, both sides start talking in the same gap. This is a human problem as much as a technical one:

  • Headsets for every remote speaker. A headset with a close microphone avoids echo and stops the browser's echo cancellation from cutting the start of words, which makes overlaps worse. Laptop speakers and microphones are the most common cause of clipped first syllables.
  • The chair names the next speaker. “Dr Amal, then Mr Karim online” removes the guesswork about who speaks next. Ask remote speakers to wait a beat after their name is called.
  • Use the raise-hand or chat for interventions. Remote participants signal before they speak instead of jumping into a pause.
  • One microphone at a time in the hall. Ask the audio technician to keep only the active microphones open, so the online audience does not hear two voices at once.
  • Brief speakers in advance. A one-page protocol sent with the joining link helps more than any setting. Our guide on how to speak in an interpreted virtual meeting gives a ready-made version.

Interpreters need the right feed, not just a fast one

In an interpreted hybrid event, overlaps hurt twice. When two people speak at once, interpreters cannot render both, and listeners on each language channel lose part of the exchange. Make sure remote interpreters hear the same clean mix as the online audience, including remote speakers and audience microphones, and give them the speaker list and running order. Interpreters also benefit from good lip-sync, because many of them watch the speaker's face to anticipate the next word.

What SmartMeet does, and what stays with the venue

SmartMeet runs in the browser with standard WebRTC audio processing: echo cancellation, noise suppression and automatic gain control in each participant's browser. Video is sent as 720p simulcast with adaptive streaming, so each viewer receives a quality that suits their connection and screen. The receiving browser re-aligns audio and video it gets from the meeting.

What no meeting platform can do is remove the delay that your cameras, switchers and capture cards add before the signal reaches the meeting computer. That offset has to be measured and corrected in the venue, as described above. In SmartMeet, listeners choose their interpretation channel from the language menu, and the host can mute everyone at once if open microphones in the online meeting start to cause overlap.

Rehearsal checklist

CheckPass when
Clap test on each cameraOffset inside the detectability window after desk delay
Clap test on the in-room screenRemote speaker's lips match their voice in the hall
Meeting computer networkWired, dedicated line, not attendee Wi-Fi
Remote speakersHeadsets on, sound checked, protocol received
Turn-takingChair names the next speaker; hand-raise tested
InterpretersHear hall, remote speakers and audience microphones clearly

Conclusion

Speaker desynchronization in hybrid events is rarely one fault. Lip-sync offset comes from the venue's video chain and is fixed by measuring it and delaying the audio to match. Overlapping speakers come from the round trip between hall and remote, and are fixed with headsets, a wired line and a chair who manages turns. Measure both at rehearsal, and your audience, speakers and interpreters will notice nothing at all.

Frequently asked questions

What is an acceptable lip-sync offset for a hybrid event?

ITU-R BT.1359 puts the threshold of detectability at roughly 45 ms with sound ahead of the picture and 125 ms with sound behind it. Aim to stay inside that window for the main screen; sound arriving early is noticed sooner than sound arriving late.

Why is the video usually behind the audio?

Cameras, video switchers, scalers, capture cards and displays all buffer frames, while a microphone signal passes through a digital mixer in a few milliseconds. The fix is to delay the audio feed by the measured offset.

How do I measure the offset without special equipment?

Clap once in front of the camera next to a live microphone, record the online meeting on a second device, and count the frames between the hands meeting and the sound starting. At 30 frames per second each frame is about 33 ms.

Why do remote speakers keep interrupting people in the hall?

Each side hears the other slightly late, so both start speaking in the same pause. Headsets, a wired connection and a chair who names the next speaker solve most of it.

Can the meeting platform fix lip-sync by itself?

The receiving browser re-aligns the audio and video it receives, but it cannot remove delay added by venue equipment before the signal reaches the meeting computer. That offset must be measured and corrected on site.

  • Hybrid Events
  • Lip Sync
  • Audio Lag
  • AV Engineering
  • Turn-Taking
  • Remote Speakers

More articles