Hi Apple Developer Community, I'm developing an eye-tracking application using ARKit's ARFaceTrackingConfiguration and ARFaceAnchor.blendShapes for gaze detection using Xcode. I'm experiencing several calibration and accuracy issues and would appreciate insights from the community. Current Implementation Using ARFaceAnchor.blendShapes (.eyeLookUpLeft, .eyeLookDownLeft, .eyeLookInLeft, .eyeLookOutLeft, etc.) Implementing custom sensitivity curves and smoothing algorithms Applying baseline correction and coordinate mapping Using quadratic regression for calibration point mapping Issues I'm Facing 1. Calibration Mismatch Red dot position doesn't align with where I'm actually looking Significant offset between intended gaze point and actual cursor position Calibration seems to drift or become inaccurate over time 2. Extreme Eye Movement Requirements Need to make exaggerated eye movements to reach screen edges/corners Natural eye movements don't translate to proportional cursor movement Difficulty reaching certain screen regions even with calibration 3. Sensitivity and Stability Issues Cursor jitters or jumps around when looking at center Too much sensitivity to micro-movements Inconsistent behavior between calibration and normal operation 4. I also noticed that tracking on calibration screen as well as tracking on reading screen works better as expected when head movement is there, but I do not want much head movement. I want tracking with normal eye movement while reading an Ebook. Primary Question: Word-Level Eye Tracking Feasibility Is word-level eye tracking (tracking gaze as users read through individual words in an ebook) technically feasible with current iPhone/iPad hardware? I understand that Apple's built-in eye tracking is primarily an accessibility feature for UI navigation. However, I'm wondering if the TrueDepth camera and ARKit's eye tracking capabilities are sufficient for: Tracking natural reading patterns (left-to-right, line-by-line progression) Detecting which specific words a user is looking at Maintaining accuracy for sustained reading sessions (15-30 minutes) Working reliably across different users and lighting conditions Questions for the Community Hardware Limitations: Are iPhone/iPad TrueDepth cameras capable of the precision needed for word-level tracking, or is this beyond current hardware capabilities? Calibration Best Practices: What calibration strategies have worked best for accurate gaze mapping? How many calibration points are typically needed? Reading-Specific Challenges: Are there particular challenges when tracking reading behavior vs. general gaze tracking? Alternative Approaches: Are there better approaches than ARKit blend shapes for this use case? Current Setup Devices: iPhone 14 Pro iOS Version: iOS 18.3 ARKit Version: Latest available Any insights, experiences, or technical guidance would be greatly appreciated. I'm particularly interested in hearing from developers who have worked on similar eye tracking applications or have experience with the limitations and capabilities of ARKit's eye tracking features. Thank you for your time and expertise!

Hi Apple Team, I’m working on a human portrait scanning application using PhotogrammetrySession, and I’ve been very impressed by the results. Thank you for building such a powerful and accessible photogrammetry solution into macOS! I do, however, have a question regarding mesh detail limitations on different Mac hardware configurations. When using PhotogrammetrySession.Request.Detail.custom and trying to set maximumPolygonCount = 1000000, I see the following log message: Clamped max poly count: 1000000 to device limit. 250000 is used. This is on an M1 Max with 32 GB RAM. I’m aware that PhotogrammetrySession.limits can report values like maximumInputImageDimension and maximumNumberOfInputImages, but I haven’t found documentation on how the maximumPolygonCount is determined, and what hardware specs influence it. Is it tied more to: • GPU performance (e.g. neural/graphics cores)? • CPU architecture? • Memory size or bandwidth? • Or is it fixed per SoC generation? I’d love to understand what kind of hardware upgrades (e.g. moving to M4 Pro or increasing RAM) could allow me to increase mesh complexity and generate more detailed models. Any insights would be greatly appreciated—and if this is covered in upcoming WWDC sessions or documentation, I’d be happy to tune in. Thanks in advance! KitCheng

I have read in the apple documentation and on forums that in order to access the camera and capture images on VisionPro, both an Entitlement and an Enterprise.license are required. I already have the Entitlement, but I don’t yet have the Enterprise.license. I would like to ask: is the Enterprise.license strictly required to gain camera access for capturing images? How can I obtain this file, and does it require an Enterprise account? Currently, my developer account is a regular Developer 99$, not an Enterprise account.

A few users have recently reported no longer being able to capture point clouds using our app, specifically on iPhone 15 Pro devices. We recently found an in-house device that exhibits this behavior and found that the confidenceMap contains only low confidence values, regardless of the environment being captured. Our app uses a higher confidence threshold; setting the threshold to a lower value produces noisy results as expected, so that is a non-viable option. Other LiDAR based apps have been tested with this device and the results are the same. No points, or noisy point clouds in apps that allow a lower confidence threshold setting. On devices that exhibit this behavior the "Displaying a point cloud using scene depth" Apple sample app can be used to visualize the issue. First reports of this new behavior occurred as early as iOS 18.4. Looking for recommendations on which team(s) at Apple to reach out to with these findings since the behavior manifests on only a small sample of devices.

Hi, When viewing a spatial photo scene on the Apple Vision Pro Photos app, you can tap on the immersive icon on the top right corner to transaction from the window presenting the image as spatial3d to an immersive photo scene with spatial3DImmersive where the window borders disappear. Could someone explain how to achieve that? I tried to do it but once I transition from spatial3d to spatial3DImmersive I can see still see a rectangle around the spatial image. Thanks.

I am trying to launch a fully immersive game from Unity on a SwiftUI view. The game is using Metal Rendering with Compositor Services. I added the unity Xcode project into the workspace, added the necessary bridge code. When I click on the button to call ufw?.showUnityWindow(), it does not start and I get the following in the console: AR session failed to start after 5 seconds. Is the app configured to use an immersive space?

Hi, since iOS 18 UnlitMaterial and ShaderGraphMaterial have the option to disable tone mapping, e.g via https://developer.apple.com/documentation/realitykit/unlitmaterial/init(applypostprocesstonemap:) Is it possible to do the same for CustomMaterial? I tried initializing a CustomMaterial based on an UnlitMaterial where tone mapping is disabled, like so: let unlitMat = UnlitMaterial(applyPostProcessToneMap: false) let customMaterial = try CustomMaterial( from: unlitMat, surfaceShader: surfaceShader, geometryModifier: geometryModifier ) but that does not seem to work. The colors of my texture still look altered in comparison to a plain UnlitMaterial or a ShaderGraphMaterial where its disabled. Any hints? Thank you!

Hi, I am in the process of implementing SharePlay into our app. The shared experience opens an Immersive Space and we set systemCoordinator.configuration.supportsGroupImmersiveSpace = true Now visionOS establishes a shared coordinate space for the immersive space. From the docs: To achieve consistent positioning of RealityKit entities across multiple devices in an immersive space during a SharePlay session There are cases where we want to position content in front of the user (independent of the shared session, and for each user individually). Normally to do that we use the transform retrieved via worldTrackingProvider.queryDeviceAnchor.originFromAnchorTransform to position content in front of the user (plus some Z Offset and smooth interpolation). This works fine in non-SharePlay instances and the device transform is where I would expect it to be but during the FaceTime call deviceAnchor.originFromAnchorTransform seems to use the shared origin of the immersive space and then I end up with a transform that might be offset. Here is a video of the issue in action: https://streamable.com/205r2p The blue rect is place using AnchorEntity(.head, trackingMode: .continuous). This works regardless of the call and the entity is always placed based on the head position. The green rect is adjusted on every frame using the transform I get from worldTrackingProvider.queryDeviceAnchor. As you can see it's offset. Is there any way I can query query this transform locally for the user during a FaceTime call? Also I would like to know if it's possible to disable this automatic entity transform syncing behavior? Setting entity.synchronization = nil results in the entity not showing up at all. https://developer.apple.com/documentation/realitykit/synchronizationcomponent Is SynchronizationComponent only relevant for the legacy MultiPeerConnectivity approach? Thank you!

Greetings. I am having this issue with a Unity Polyspatial VisionOS app. We have our main Bounded Volume for our app. We have other Native UI windows that appear when we interact with objects in our Bounded Volume. If a user closes our main Bounded Volume...sometimes it quits the app. Sometimes it doesn't. If we go back to the home screen and reopen the app, our main Bounded Volume doesn't always appear, and just the Native UI windows we left open are visible. But, we can sometimes still hear sounds that are playing in our Bounded Volume. What solutions are there to make sure our Bounded Volume always appears when the app is open?

The texture baking section of the WWDC25 session "Optimize your custom environments for visionOS" (https://youtu.be/RELnRZmb02c?t=1485) moves very quickly and leaves a lot unexplained. Has anyone worked through this part of the toolkit in practice and can speak to what's actually going on, particularly around projection baking and how it addresses the reprojection artifacts the presenter briefly mentions? Thankyou

The initial startup of visionOS 26 after install is glacially slow.

While using apple's vision pro, we noticed that we can continue to use the visionOS keyboard when we no longer actually see it in passthrough. In other words, when we focus on a field to type, visionOS displays the keyboard for us in such a way that we actually see it. Then, we noticed if we look away a little bit, either up, or down, or left, or right, in such a way that the keyboard is no longer visible by us in the passthrough, the keyboard still remains responsive to taps from our fingers at the location where it is. It seems the keyboard remains functional and responsive to taps even though we can no longer observe/see it. We are trying to figure out how to implement similar functionality in our app whereby the user can continue to manipulate a 3d entity when the user can no longer actually observe it in passthrough (like the visionOS keyboard appears to allow). I assume the visionOS keyboard has this functionality thanks to the downward facing sensors on the hardware that allow hand tracking even though the hands can no longer be observed by the user. That is likely how we can rest our hands on our lap is still be able to interact with visionOS. How can we implement a similar functionality for 3D entities? Is there a way to tap in, or to allow hand tracking, from those toward facing cameras? Is it possible to manipulate a 3D entity when it is no longer observed by the user for example when they shift their attention somewhere else in the field of vision? How does the visionOS keyboard achieve this?

Hi, I would like clarification on whether the new hover effects feature introduced in vision os 26 supported pinch gestures through the psvr 2 controllers. In your sample application, I found that this was not working. Pulling the trigger on the controller whilst looking at the 3d object did not activate the hover effect spatial event in the sample application. (The object is showing the highlight though), only pinch clicking with my fingers seem to be registering/triggering the spatial event. I am using Vision OS 26.3 This is inconsistent with how the psvr2 controller behaves on swift ui views and ui view elements, where the trigger press does count as a button click. The sample I used was this one: https://developer.apple.com/documentation/compositorservices/rendering_hover_effects_in_metal_immersive_apps

Use case proposal: immersive mental visualisation and manifestation experiences on visionOS I'd like to open a discussion around a use case I believe is currently missing from the visionOS ecosystem

I am creating a vision pro app with a 3D model, it has a mesh hierarchy of head, hands, feet etc. I want the character to look towards the camera, but am not able to access head of character through sceneKit nor reality kit. when I try to print names of the child meshes, it only prints till the character, it does iterate through all the body parts. Can anyone help?

Anyone could share ideas or nodes setup to implement a gaussian blur on shader graph material, with a blur size parameter? Thanks!

I'm placing sphere at finger tip and updating its position as hand move. Finger joint tracking functions correctly, but I’ve observed noticeable latency in hand tracking updates whenever a UITextView becomes active. This lag happens intermittently during app usage, lasting about 5–10 seconds, after which the latency disappears and the sphere starts following the finger joints immediately. When I open the immersive space for the first time, the profiler shows a large performance spike upto 328%. After that, it stabilizes and runs smoothly. Note: I don’t observe any lag when CPU usage spikes to 300% (upon immersive view load) yet the lag still occurs even when CPU usage remains below 100%. I’m using the following code for hand tracking: private func processHandTrackingUpdates() async { for await update in handTracking.anchorUpdates { let handAnchor = update.anchor if handAnchor.isTracked { switch handAnchor.chirality { case .left: leftHandAnchor = handAnchor updateHandJoints(for: handAnchor, with: leftHandJointEntities) case .right: rightHandAnchor = handAnchor updateHandJoints(for: handAnchor, with: rightHandJointEntities) } } else { switch handAnchor.chirality { case .left: leftHandAnchor = nil hideAllJoints(in: leftHandJointEntities) case .right: rightHandAnchor = nil hideAllJoints(in: rightHandJointEntities) } } await MainActor.run { handTrackingData.processNewHandAnchors( leftHand: self.leftHandAnchor, rightHand: self.rightHandAnchor ) } } } And here’s the function I’m using to update the joint positions: private func updateHandJoints( for handAnchor: HandAnchor, with jointEntities: [HandSkeleton.JointName: Entity] ) { guard handAnchor.isTracked else { hideAllJoints(in: jointEntities) return } // Check if the little finger tip and intermediate base are both tracked. if let tipJoint = handAnchor.handSkeleton?.joint(.littleFingerTip), let intermediateBaseJoint = handAnchor.handSkeleton?.joint(.littleFingerIntermediateTip), tipJoint.isTracked, intermediateBaseJoint.isTracked, let pinkySphere = jointEntities[.littleFingerTip] { // Convert joint transforms to world space. let tipTransform = handAnchor.originFromAnchorTransform * tipJoint.anchorFromJointTransform let intermediateBaseTransform = handAnchor.originFromAnchorTransform * intermediateBaseJoint.anchorFromJointTransform // Extract positions from the transforms. let tipPosition = SIMD3<Float>(tipTransform.columns.3.x, tipTransform.columns.3.y, tipTransform.columns.3.z) let intermediateBasePosition = SIMD3<Float>(intermediateBaseTransform.columns.3.x, intermediateBaseTransform.columns.3.y, intermediateBaseTransform.columns.3.z) // Calculate the midpoint. let midpointPosition = (tipPosition + intermediateBasePosition) / 2.0 // Position the sphere at the midpoint and make it visible. pinkySphere.isEnabled = true pinkySphere.transform.translation = midpointPosition } else { // If either joint is not tracked, hide the sphere. jointEntities[.littleFingerTip]?.isEnabled = false } // Update the positions of all other hand joint spheres. for (jointName, entity) in jointEntities { if jointName == .littleFingerTip { // Already handled the pinky above. continue } guard let joint = handAnchor.handSkeleton?.joint(jointName), joint.isTracked else { entity.isEnabled = false continue } entity.isEnabled = true let jointTransform = handAnchor.originFromAnchorTransform * joint.anchorFromJointTransform entity.transform.translation = SIMD3<Float>(jointTransform.columns.3.x, jointTransform.columns.3.y, jointTransform.columns.3.z) } } I’ve attached both a profiler trace and a video recording from Vision Pro that clearly demonstrate the issue. Profiler: https://drive.google.com/file/d/1fDWyGj_fgxud2ngkGH_IVmuH_kO-z0XZ Vision Pro Recordings: https://drive.google.com/file/d/17qo3U9ivwYBsbaSm26fjaOokkJApbkz- https://drive.google.com/file/d/1LxTxgudMvWDhOqKVuhc3QaHfY_1x8iA0 Has anyone else experienced this behavior? My thought is that there might be some background calculations happening at the OS level causing this latency. Any guidance would be greatly appreciated. Thanks!

I have a question about Apple’s preinstalled visionOS app “Encounter Dinosaurs.” In this app, the dinosaurs are displayed over the real-world background, but the PhysicallyBasedMaterial (PBM) in RealityKit doesn’t appear to respond to the actual brightness of the environment. Even when I change the lighting in the room, the dinosaurs’ brightness and shading remain almost the same. If this behavior is intentional — for example, if the app disables real-world lighting influence or uses a fixed lighting setup — could someone explain how and why it’s implemented that way?

Hi all, I'm running into an issue with an app that previously worked fine on device using visionOS 2.0. After updating to visionOS 26, the same code runs fine in the simulator but crashes on the device with the following error: -[MTLDebugComputeCommandEncoder _validateThreadsPerThreadgroup:]:1330: failed assertion `(threadsPerThreadgroup.width(32) * threadsPerThreadgroup.height(32) * threadsPerThreadgroup.depth(1))(1024) must be <= 832. (kernel threadgroup size limit)` Is there any documented way to check or increase the allowed threadsPerThreadgroup size on Apple Vision Pro? Or any recommended workaround for this regression? Thanks in advance!

Hi, I have been using RealityRenderer to render scenes in MacOS as spatial videos and view it in Vision Pro and it is awesome. I understand that it uses PerspectiveCamera to render. I wanted to know what is the default FOV for this camera and how much can we push it? I want to ideally render a scene with 180 degrees of fov. Thanks