By Apple released visionOS 26.4 this week, and the update that matters most is one you hear, not see. Spatial Audio on Vision Pro finally works the way it should have from the start.
Since launch, every sound in a visionOS app played from a single point: wherever the app's first window happened to open. Three windows spread across your living room, and all of them spoke from the same spot. The video call on your left, the notification on your right, the game straight ahead, all collapsed into one invisible speaker. You felt something was off before you could explain what.
Per-window audio changes everything it should
The Spatial Audio Experience API, first announced at WWDC 2025, ships to all users with this update. Each sound source can now originate from its own window or volume. Audio tracks visual placement. A FaceTime voice comes from the FaceTime window. A notification pings from the direction of the notification panel. A game's ambient soundtrack wraps around you from the volume it actually occupies.
Why did it take this long? As Next Reality explained, the single-anchor model was inherited from AudioToolbox and AVFoundation, frameworks built before multi-window spatial environments existed. When a visionOS app launched, the system needed a spatial reference point for audio and defaulted to the first window. That worked fine when most apps used a single window. As Apple pushed multiwindow productivity, the inherited limitation became impossible to ignore.
Sounds can also move between scenes without cutting out, which Apple demonstrated at WWDC 2025 as a deliberate design goal. For narrative apps or games that transition between environments, audio continuity during a scene shift is the difference between polished and rough.
This is a platform-model change, not a hardware upgrade. Apple hasn't disclosed whether latency or processing overhead shifted, or how many simultaneous sources the system supports. But the behavioral change is clear: audio now comes from where things are, and that alone makes multiwindow setups feel like spatial computing instead of a desktop with extra steps.
Room-aware acoustic caching
The second audio improvement is subtler but clever. According to Apple's release notes, visionOS 26.4 makes Spatial Audio start faster in familiar spaces by remembering the acoustic properties of rooms you've been in before.
The headset already uses what Apple calls Audio Ray Tracing, scanning the features and materials of a space to match sound to the physical environment. That process took time each session. Now the device caches acoustic profiles of rooms it recognizes, so calibration in your home office or living room happens near-instantly instead of requiring a fresh scan every time you put on the headset.
As 9to5Mac put it, the device is now "capable of recognizing your environment and remembering how best to deliver Spatial Audio." It's the kind of invisible engineering that you only notice when it stops getting in the way.
What developers can build now
The per-window audio API opens design possibilities that weren't available before. A multiwindow productivity app can give each panel its own spatial voice. A game distributing sound sources across a mixed reality environment can now track objects moving through physical space instead of collapsing everything to one anchor. Shared-space experiences, where multiple users occupy the same room, benefit from audio that originates from the correct visual position rather than creating spatial confusion.
Scene transitions gain a specific capability worth noting. According to UploadVR's coverage, sounds can move between scenes without positional cuts, which Apple designed as an explicit goal. For any app that moves users between spatial environments, audio continuity during the shift separates work that feels finished from work that feels like a demo.
The API operates across both AudioToolbox and AVFoundation. Existing apps don't need a full rewrite; they need to adopt the new spatialization anchor model.
Why this matters more than it looks
I've worn the headset through enough sessions to know what subtle wrongness feels like. You arrange your windows, you set up your space, and then everything sounds like it's coming from one direction. You can't quite place the problem, but your brain registers the mismatch between where things look and where they sound. That gap erodes the illusion that spatial computing is supposed to create.
visionOS 26.4 closes that gap. Per-window spatialization and room-aware acoustic caching are infrastructure fixes that don't make flashy keynote demos but make the headset worth wearing for longer stretches. The audio now matches the visual layout. Your living room sounds like your living room on the second session, not just the tenth.
If you've been on the fence about whether Apple is serious about this platform, this is the kind of update that shows they're iterating where it counts, not where it's most visible.
Ren Wilder covers mixed reality for The Daily Vibe.
