Video Call from the Future with Spatial Audio for Zoom, MS Teams

Remote meetings are no longer the exception, and most of us now spend more hours per week on Teams, Zoom or Meet than in any physical meeting room. Yet if you join from a laptop or a phone, or sit a metre away from your microphone, what the other side hears is usually a low-quality, centrally mixed stream.

That matters more than it sounds. Group conversation runs on small timing, level and spectral differences between your two ears, and those are exactly the cues a platform throws away when it collapses several microphone feeds into one channel.

Below: what gets lost, what spatial audio puts back, which platforms already ship it, and what you can change on your own machine this afternoon.

Why video calls sound flat

A real conversation happens around a circle. People sit at angles, voices arrive from different directions, and your brain separates the talkers instantly without you noticing any effort.

Remote meeting stacks do the opposite. Multiple microphone feeds are collapsed into a single, centrally positioned stream. On top of that the platform applies noise suppression, automatic gain control and single-speaker optimisation, all of which flatten the spectral and level differences that were carrying the spatial information.

How your ears actually locate a voice

Three mechanisms do the work, and a mono call defeats all three.

Interaural time differences, ITD, are the slight difference in arrival time between your two ears. A voice from your left reaches the left ear a fraction of a millisecond earlier, and your auditory system is astonishingly good at reading that delay.

Interaural level differences, ILD, are the difference in intensity between the ears. Your head casts an acoustic shadow, so the far ear hears less. The effect is strongest at high frequencies, where the head blocks sound most effectively.

Spectral cues are the frequency colouring that your head, pinnae and torso impose on a sound before it reaches the eardrum. Higher frequencies are filtered most, and your brain reads those changes to work out elevation and whether a source is in front of or behind you. This is the part that carries the vertical and sagittal planes, where timing and level differences alone tell you almost nothing.

Spectral and level cues carry most of the load in complex environments, exactly where binaural timing cues get unreliable. The brain also gets better at combining them with experience, which is why a format that strips them out is worse than it looks on paper.

In-head localisation: why remote voices end up inside your skull

When the same, or a tightly correlated, signal reaches both ears at once with no meaningful ITD, no ILD and no spectral difference, your brain has nowhere to put the sound. It concludes the source must be inside your head.

Audio engineers call this in-head localisation, and it is the normal outcome rather than an edge case. On headphones you hear other people’s voices from somewhere between your ears. On laptop speakers you get the milder version: a single indistinct blob in front of you.

The consequence is cognitive, not acoustic. Once the spatial separation is gone, you have to do extra work to segregate voices, identify who is speaking and follow the turn-taking. The effort compounds across a working day.

Externalisation: what spatial audio actually does

The opposite of in-head localisation is externalisation, and this is the part I promised to come back to.

Over headphones it works by treating each participant as a separate audio object rather than mixing everyone together first. For each object the renderer delays and attenuates one ear slightly, shapes the tone as if the sound had bounced off your head and outer ear, and adds small room reflections. Your brain reads those cues and places the voice outside your head.

The output is still an ordinary two-channel signal. Only the percept changes, and it changes completely. It suddenly sounds as though you are not wearing headphones at all, because you are localising the voices around you rather than inside you. Even most laptop stereo speakers can convey some of this.

Worth stressing: this is software. You do not need a spatial microphone, a special headset or any new hardware. An ordinary laptop microphone and a pair of wired headphones are enough.

In real life, people don’t sit on a single chair in front of you. They sit around you. Spatial audio in teleconferencing puts them back where they belong.

Where spatial audio already works today

Platform support has changed substantially since this article first appeared.

Does Zoom have spatial audio?

Yes. Zoom introduced spatial audio for meetings and webinars, positioning each participant’s voice in the stereo field according to their place in Gallery or Immersive view.

It requires the Zoom desktop app from version 6.0.10 on Windows, macOS or Linux. It is limited to Gallery, Side-by-side Gallery and Immersive views, it applies only to participants’ microphone audio and not to audio inside a shared screen, and it does not work over Bluetooth. Zoom Rooms support it as well when fitted with stereo loudspeakers.

There is a quick way to check whether your output device qualifies: enable spatial audio in the Audio settings, then use Test Speaker. With spatial audio active the ringtone plays five times, each from a different position. If you hear one static ringtone, your device is not supported.

Separately from spatial audio, Zoom also offers stereo and high-fidelity modes. Those do not place voices in space, but they do raise the quality of what gets transmitted, and they let you send two-channel material. In practice I use this regularly to demonstrate spatial audio examples over a call, for a reason hardly anyone knows: you can transmit in stereo twice over, once through the microphone and again through screen sharing. I know of no other conferencing software that manages both, neither Teams nor Discord nor FaceTime. It is how binaural samples get played at events like this one.

What is spatial audio in Teams?

Teams places each participant’s voice at their position in the gallery. Like Zoom, it needs USB-wired stereo headphones or speakers, or your device’s built-in stereo speakers. Bluetooth audio devices are not supported, because they drop into a phone mode that only allows data-reduced mono playback.

The meeting has to have more than two participants in gallery view for the effect to engage, and Teams switches it off automatically when network bandwidth or computer memory runs low.

To turn it on: open Settings and more, then Devices, select your compatible device under Speakers and switch on the toggle for spatial audio playback. The setting persists across meetings. You can also set it per meeting through Device Settings before joining from the calendar.

Two limitations are worth knowing before you enable it for a whole organisation. In meetings with live interpretation, spatial audio makes the original and translated audio play at the same volume, which defeats the purpose of interpretation. And with music mode enabled you lose the high-fidelity full-band path. In both cases you have to turn spatial audio off and rejoin.

Official guide from Microsoft

Apple FaceTime

Apple added spatial audio to FaceTime with iOS 15, so voices are heard from the direction of the participant’s tile on screen, alongside more aggressive background noise filtering. The filtering matters more than it might appear: spatial rendering works best when each voice arrives as isolated as possible, because any leftover room noise gets spatialised too.

Dolby OptiView

Dolby OptiView, formerly the Dolby.io Communications API, brings spatial audio into conferencing at the API level. Participants are placed in a shared 3D audio space and hear each other from their respective positions. When someone speaks, their circle lights up on the page and the sound comes from that part of the interface. For developers building their own conferencing product rather than configuring someone else’s, this is the relevant layer.

Earlier experiments pointed the same way. Clubhouse rolled out spatial audio to iOS in 2022 and ran into the stereo problem that every implementation hits: a stereo source is far harder to place than a mono one. Their workaround was to derive mono sources from the left and right channels, keeping some stereo richness while gaining spatial depth.

A note on terminology

Vendors do not agree on a name. Microsoft often says immersive audio, Zoom says spatial audio, others say object-based or binaural. Underneath, the pattern is the same in each case: keep each speaker separate, position them, render binaurally, add subtle room cues.

That matters when you are comparing products or writing a specification. Judge the mechanism, not the marketing label.

Benefit 1 – Telling voices apart

One-on-one calls are not really the problem. With a single other person your brain leans on context and on the video, so spatial cues add only a little.

The moment a third participant joins, the demands change. Overlapping speech, similar pitch or timbre and fast turn-taking make it hard to tell who said what, and it gets worse with every additional person. That is why both Teams and Zoom only engage spatial audio above two participants. It is not an arbitrary threshold.

Distribute each voice across the 3D audio space the way reality does, and the separation comes back. Strictly speaking even plain stereo helps, as the demo by High Fidelity founder Philip Rosedale makes audible. But that example also shows the limit: there is no reverberation, so you get left-right placement rather than true externalisation, and the voices sit uncomfortably close to the ear.

These are not hypothetical benefits. Spatial separation enables spatial unmasking, the same psychoacoustic effect that lets you follow one voice in a noisy room while ignoring the others, better known as the cocktail party effect.

Benefit 2 – Head tracking

Small head movements sharpen localisation. In everyday life you turn your head slightly and unconsciously when you want to pin down a sound, because changing the angle changes both the arrival time and the level difference between your ears, and the brain uses that change as extra evidence.

Bringing those movements into a video call sounds like a lot of engineering, but the technology already exists in several forms.

Head tracking via headphones

Apple ships head-tracking headphones in the AirPods Pro and AirPods Max, Samsung in the Galaxy Buds Pro, to name the two best-known. You can also attach external head trackers to ordinary headphones. There is more detail in this article.

Capture head rotation via the webcam

Not everyone owns such devices, which makes webcam-based head tracking the interesting option for desktop use. Face detection supplies the head orientation, and the 3D sound field is adapted to it in real time.

For spatialised video calls this is close to a free lunch. The call already requires a camera, so no hardware purchase is involved, and the spatialisation can run in the browser. The team at atmoky is furthest ahead here.

Video call without video?

Follow the thought and the video becomes less essential, at least for telling people apart. A spatialised telephone call would already convey a very different sense of the other person’s presence, which is a psychological effect as much as an acoustic one, especially in a plug-and-play system.

Several bird-view meeting spaces work on this principle, among them Gather Town, SpatialChat and the already mentioned High Fidelity.

Benefit 3 – Zoom fatigue and listening effort

Anyone with several video conferences a day knows the feeling of being drained afterwards without having done much. The phenomenon is called Zoom fatigue, and it is not necessarily about the content of the meeting.

Mono playback demands more processing from your brain. Every voice arrives without directional information, so the brain spends effort separating and assigning them, effort that a face-to-face conversation simply does not require. Rennies and Kidd measured exactly this in the Journal of the Acoustical Society of America: binaural listening improves both speech intelligibility and listening effort, and the two do not always move together (paper).

Restore the directional scene and you hand that work back to the auditory system, where it runs for free. Meetings get less tiring and, over a day, measurably more productive. The atmoky team is pursuing exactly this in web meetings, building a natural acoustic scene to raise speech intelligibility and exploit spatial unmasking properly.

Why this matters most for people with hearing loss

For listeners with normal hearing, a flat mono mix is tiring. For people with hearing loss it can be the difference between following a meeting and not following it.

Spatial release from masking is one of the mechanisms that degrades with hearing loss in the first place. Strip the spatial cues out of the call and you have taken away a coping strategy that was already under strain.

Per-speaker rendering gives it back, and the gain is larger for this group than for anyone else in the meeting.

If your organisation has an accessibility policy that covers captions, spatial audio belongs in the same conversation. It costs nothing to enable and it is the same feature that makes the meeting better for everyone else.

Get the playback right before you worry about the format

The best spatial rendering in the world cannot survive bad playback. Most of what follows is free.

Headphone etiquette

Ask everyone to wear headphones. Laptop speakers bleed into local microphones, and while meeting clients try to suppress that bleed, they frequently end up muting useful speech or introducing artefacts instead. Headphones remove the speaker bleed, cut echo and background noise, and preserve speech clarity for both sides. It is a small piece of etiquette that removes a whole class of problems.

If you personally struggle to follow calls, a decent pair of headphones is the first thing to change, because they improve intelligibility regardless of what the platform does.

Which headphones

Closed-back over-ear is the best general choice: larger drivers, fuller tone, and passive isolation. That combination is why they dominate in professional audio.

On-ear is the compact compromise. Lighter and more portable, reasonable speech clarity, less isolation.

In-ear earbuds are the smallest and most practical option. Modern high-end models deliver surprisingly good intelligibility, provided they actually seal.

In all three cases, wired usually sounds best. Many call clients drop a Bluetooth device into a phone-mode or low-bit codec as soon as it is used as a microphone, which costs you fidelity and, as noted above, disables spatial audio in both Teams and Zoom.

Client settings and the 60-second A/B check

Every meeting client treats sound differently, so the only reliable method is to test rather than to assume. Toggle any high-quality, original sound or hi-fi option. Switch noise suppression between Low and Off. Compare wired against Bluetooth briefly.

Then run short A/B checks of 30 to 60 seconds. Sometimes noise suppression helps a great deal. Sometimes turning down dereverberation keeps voices fuller and easier to follow, and sometimes it strips away too much of the original dialogue. It is subjective, so listen and keep whichever combination makes speech clearest for your room and your hearing.

Why the vocals in some webinar clips are so hard to follow

A question I get often: why is the voice track in some short videos during webinars or online training much harder to follow than in others?

The answer is frequency overlap plus dynamics. Music fills almost the entire audible band, roughly 20 Hz to 20 kHz, while speech occupies around 100 Hz to 8 kHz. They compete for the same frequencies and mask each other.

On top of that, music is usually mixed and dynamically compressed to hold a fairly constant level, whereas speech is highly dynamic and fluctuates. A steady musical bed masks a variable vocal track continuously rather than in gaps, and for listeners with hearing loss the effect is considerably worse. If you produce this material, ducking the bed under speech is worth more than any amount of vocal EQ.

Challenges

What you do not need is an expensive 3D microphone, although there is a nice overview here if you are curious. A normal mono microphone on a headset, or the one already built into your laptop, is enough. Spatialisation happens in software: metadata is added to the audio stream and the renderer calculates in real time how the virtual scene should sound.

The real problem is what happens before that. Audible artefacts from data reduction during transmission are still common, and every noise-suppression stage is a chance to lose speech. NVIDIA’s RTX Voice showed how far noise cancellation can go, which is doubly useful here: you do not want a virtual sound source that also carries laptop fan noise and keyboard clatter, you want clean speech placed in space.

The other open issue is the gap between consumer and room hardware. A participant on wired headphones gets the full effect. The same meeting joined from a conference room with a ceiling array, or from a phone on Bluetooth, does not. Hybrid meetings therefore deliver two quite different experiences from the same stream, and that is not yet solved.

Conclusion and solutions

Audio was the last part of video conferencing to get serious attention, and it is the part with the most headroom left. Remote meetings are permanent now, so good audio is no longer a nicety.

Restoring the spatial cues costs less listening effort, and people work out who is speaking faster. For participants with hearing loss it can decide whether they follow the meeting at all. Major vendors are already shipping the capability.

If you run an organisation, three steps cost almost nothing: enable the spatial or immersive features you already pay for, publish simple playback guidance, and run quick A/B checks to find what actually works for your audience. Those small steps turn a flattened, attention-hungry conference into something that behaves more like the real world.

With my colleagues at atmoky I can implement this directly for web meetings, virtual interaction and video calls.

A version of this article appeared in Hearing Matters, the magazine of Hearing Matters Australia, August/September 2026 edition.

Let’s talk about your project

Want to use 3D audio in a real project, not just understand it in theory?