Can we spatialise VoiceOver speech on macOS using AirPods Max?

By VivekP, 6 November, 2026

Forum
Mac

Hello everyone,

I've been wondering about something for a while, and I thought I would put the idea before the AppleVis community, particularly those who have experience with macOS audio programming, Core Audio, and accessibility development.

We know that Apple's Spatial Audio technology, particularly with AirPods Max and compatible AirPods, can create the impression that sound originates from a position outside the listener's head. With dynamic head tracking, the apparent sound source can even remain anchored in space as the listener moves their head.

Now, here's my question: Could we somehow apply similar spatial processing to VoiceOver's speech output on macOS?

I'm not referring to the existing positional audio option in VoiceOver Utility, which provides spatial cues for interface sounds. Nor am I referring to enabling Spatial Audio through Control Centre. I'm interested specifically in the synthesised speech produced by VoiceOver.

Imagine having VoiceOver's voice appear to originate from a fixed point in front of you, perhaps somewhere near the MacBook's screen, rather than hearing it as a conventional headphone-centred voice.

At this stage, I'm not proposing that different interface elements should occupy different positions in a three-dimensional auditory environment, although that might be an interesting direction for future development. My immediate interest is much simpler: making the VoiceOver voice itself sound spatially externalised.

I wonder whether this could be accomplished through a combination of virtual audio routing and binaural processing.

For example, could VoiceOver's audio output be routed through a virtual Core Audio device, processed using an HRTF-based spatial audio engine, and then delivered to AirPods Max? Could such an arrangement work with sufficiently low latency for ordinary screen-reader navigation?

A more ambitious possibility would involve using the headphones' motion data to keep the virtual voice anchored to a fixed location as the listener turns their head.

I'm curious about several technical questions:

1. Is VoiceOver's speech output accessible to third-party audio-processing software through Core Audio or another macOS mechanism?
2. Are there existing virtual audio devices, audio processors, or spatialisation tools that might accomplish this without requiring a completely new application?
3. Could Apple's headphone-motion APIs be used for dynamic head tracking in such an arrangement?
4. What technical or accessibility limitations might make this impractical, particularly regarding latency, speech intelligibility, and reliable audio routing?

I realise that Apple may not expose VoiceOver's audio stream in a way that makes this straightforward. I'm also aware that ordinary stereo widening or adding reverberation is not quite the same thing as convincing binaural spatialisation.

Nevertheless, given what modern spatial audio processing can accomplish, I wonder whether this is technically feasible.

I'm using an Apple Silicon M5 MacBook Air with AirPods Max second generation and AirPods Pro second generation. I'd be particularly interested to hear from developers, audio programmers, or anyone who has experimented with virtual audio devices and spatial sound on macOS.

Perhaps someone has already tried something along these lines. Or perhaps it might inspire a small experimental project.

I'd be delighted to hear your thoughts.

Options

Comments

By Levi Gobin on Friday, November 6, 2026 - 11:04

Member of the AppleVis Editorial Team

logic pros the only way I know how to do this currently, you may be able to do it another way. What I would do is download an app like Loopback or use black hole to create an audio device which feeds VoiceOver to Logic Pro. Then in logic, enable Dolby Atmos and voiceOver will be spatialized. You can then enable headtracking and the sound will stay fixed to the computer while positional audio in VoiceOver itself continues to work correctly. If you really wanted to, you could write a custom program which could pan VoiceOver speech and sounds based on the currently focused item, but any app that pulls from the VoiceOver cursor regularly does add slowdown. but if all you want is VoiceOver to be virtualized, logic is a good option.

Another good option I have not tested in years is an app called boom 3-D. I have not tested it in years, but it could also virtualize spatial audio, but it does not have head tracking.

Thank you, Levi! This is precisely the sort of technical suggestion I was hoping for.

I hadn't considered using Logic Pro's Dolby Atmos renderer for live VoiceOver output. Have you actually tried routing VoiceOver through Logic in this way, particularly with AirPods Max and head tracking enabled?

I'm especially interested in whether the additional processing introduces any perceptible latency during rapid VoiceOver navigation, and whether the audio routing remains reliable.

I've also come across Orbit Monitor, which apparently supports virtual audio input and head tracking with AirPods Max. I wonder whether that might offer a lighter-weight alternative to keeping Logic Pro running.

Thanks again for the very promising suggestion!

By Choom on Friday, November 6, 2026 - 19:06

I'm not sure how much you have experience head position tracking, but it only works directionally, meaning that if you move sideways relative to the position of the sound source, the audio does not move to accommodate the new relative orientation.

As for your questions, while it is possible to route VoiceOver audio specifically, as VoiceOver itself provides an option to use a non-default audio device for output, injecting the audio back into the default output always requires an application. As for spatial audio, I think that it only requires adding more than two non-LFE audio channels to the audio source, but I've never messed with this at such a low level myself, because the functionality is also made available at a much higher level by AVFoundation's / AVAudio's audio engine's positional audio.

Essentially you can create virtual audio devices using either CoreAudio's plug-ins or actual DriverKit audio drivers, but audio devices get audio already mixed by CoreAudio's Hardware Abstraction Layer, so you cannot apply effects to specific CoreAudio clients using virtual devices, however you can route VoiceOver audio through a virtual loopback device like the one created by Blackhole, and then have an application add an extra silent non-LFE channel to the stream in a surround configuration where the two stereo channels are at the front and the extra silent channel is at the back, and then play that back on the default output device. This should be enough to enable positional audio with head position tracking, but again,I've never done this myself, I'm merely theorizing based on my hands-on experience with CoreAudio.

As for latency, every hop adds some of it since the audio streams are buffered in chunks of 512 frames if I recall correctly, which amounts to just over a millisecond of delay on a 48KHz stream.

Personally I don't think that this idea is worth any trouble, because you can just as well remove your AirPods and listen to the audio coming out of your Mac's speakers for a true to life immersive positional audio experience.