We build augmented reality experiences, visionOS spatial apps, and computer vision systems that run entirely on-device. From ARKit's first release to Apple Vision Pro on launch day, Misterios has been shipping production AR to millions of users.
visionOS introduces an entirely new computing paradigm. Windows float in your room. Objects have depth and volume. Interfaces respond to where you look and what your hands do. We build for this platform natively, using the full depth of Apple's spatial frameworks to create experiences that feel like they belong in the real world.
visionOS apps are built around three presentation modes: windows, volumes, and immersive spaces. We design and implement all three. Windows provide familiar 2D interfaces that float at comfortable distances. Volumes render 3D objects users can walk around and inspect from any angle. Full immersive spaces replace the entire environment, placing users inside your content. We compose these modes using SwiftUI's native scene types, managing transitions, ornaments, and spatial layout so your app feels considered rather than ported.
Every 3D object in a visionOS app passes through RealityKit. We work with USDZ assets authored in Reality Composer Pro, applying physically based rendering materials, particle effects, skeletal animations, and spatial audio. Our pipeline handles the full path from designer handoff to optimised runtime asset, ensuring models render at native resolution on M2 silicon without dropped frames. We build custom RealityKit systems and components for behaviours that go beyond what off-the-shelf tools provide, including physics-driven interactions, procedural geometry, and dynamic material swapping.
Spatial input on Vision Pro relies on hand and eye tracking rather than controllers. We implement custom gesture recognisers that go beyond the standard pinch-and-drag: two-handed resize, rotational gestures, flick-to-dismiss, and precision placement. Gaze-driven interactions let users highlight, select, and trigger actions simply by looking, while hand tracking provides the confirmation layer. We calibrate interaction zones and hover states so controls feel responsive without causing fatigue during extended sessions.
Collaborative spatial experiences require synchronising 3D state across devices in real time. We build SharePlay integrations that let multiple Vision Pro users occupy the same virtual workspace, seeing each other's cursors, annotations, and object manipulations with sub-second latency. This applies to collaborative 3D design review, shared data visualisation, spatial whiteboarding, and multiplayer training simulations. We handle the networking, conflict resolution, and spatial coordinate alignment so the experience feels seamless.
ARKit remains the most capable mobile AR framework available. We have been building with it since its first release in 2017, and we have shipped every major feature Apple has added since: world tracking, image detection, body tracking, face mesh, scene reconstruction, LiDAR integration, location anchors, and object capture. Here is what that looks like in production.
Real-time tattoo visualisation on skin requires body tracking with perspective-correct rendering that follows the contours of the user's body as they move. We build this using ARKit's body anchor detection combined with custom Metal shaders that warp 2D designs onto the 3D body mesh, accounting for skin curvature, lighting conditions, and camera movement. The result is a tattoo preview that looks photorealistic at 60fps and works across different body types and skin tones.
Placing a piece of furniture in your living room before you buy it requires millimetre-accurate surface detection and stable world tracking. We use LiDAR-enhanced scene understanding on supported devices to anchor 3D objects to real-world surfaces with physical accuracy, handling occlusion so virtual objects disappear behind real furniture. On non-LiDAR devices, we fall back to plane detection with refined hit-testing to maintain a quality experience across the entire iPhone lineup.
Face-based AR effects require real-time face mesh deformation, blend shape tracking, and custom rendering at 60fps with zero perceptible latency. We implement these using ARKit's face tracking combined with Metal shaders for effects that go beyond what standard RealityKit materials support, including dynamic skin retouching, animated overlays, colour grading, and physics-driven accessories that react to facial expressions and head movement.
Geo-anchored AR content pins virtual objects to real-world GPS coordinates so they persist across sessions and appear for any user who visits the same location. We combine ARKit's location anchors with MapKit and custom coordinate systems to deliver location-based experiences for retail, tourism, navigation, and outdoor gaming, handling the inherent challenges of GPS drift, compass calibration, and varying environmental conditions.
The most powerful AR experiences combine real-time rendering with on-device machine learning. Every inference runs locally on Apple Neural Engine and GPU, meaning zero cloud latency, full offline capability, and complete user privacy. We build the full pipeline from data collection through model training to production deployment.
We integrate Apple's Vision framework and custom Core ML models for real-time object detection, image classification, text recognition (OCR), barcode and QR scanning, hand pose estimation, and human action recognition. Our models are optimised for Apple Neural Engine execution using quantisation and pruning, keeping inference under 10ms per frame so the camera feed stays at a locked 60fps. We use Create ML for rapid prototyping and custom training pipelines with PyTorch or TensorFlow when the task requires architecture-level control.
Production video applications often require multi-camera capture, real-time frame analysis, custom compositing, and hardware-accelerated encoding, all running simultaneously without dropping frames. We build these pipelines using AVFoundation for capture and playback, Metal compute shaders for per-frame image processing, and VideoToolbox for hardware encoding. This architecture powers applications ranging from multi-angle sports analysis to real-time quality inspection on manufacturing lines.
Millions of downloads across iOS and Android. Real-time tattoo visualisation using ARKit body tracking with perspective-correct rendering on skin. Custom Metal shaders handle lighting adaptation and skin-tone matching. Users can browse a catalogue, place any design on their body, and see it move naturally as they rotate in front of the camera.
Built for a Fortune 500 manufacturer. Computer vision tracks assembly line workers' hand positions and tool usage in real time, providing instant corrective feedback during training sessions. Custom Core ML models trained on proprietary datasets detect 40+ distinct assembly actions. Deployed across multiple facilities, resulting in 77% fewer safety incidents during the first year of operation.
Simultaneous recording from multiple iPhone cameras with synchronised timestamps, real-time preview compositing via Metal, and hardware-accelerated H.265 encoding. The system handles frame-perfect alignment across sensors with different focal lengths, enabling picture-in-picture, split-screen, and director-mode workflows for professional content creation.
A visionOS shopping experience where products appear as life-size 3D models in the user's physical space. Customers walk around furniture, inspect material textures at close range, and place items in their actual room using spatial anchoring. Built with RealityKit and SwiftUI, the app supports real-time configuration changes — fabric, colour, dimensions — with instant re-rendering.
We work across the full breadth of Apple's spatial and vision frameworks, plus the GPU and ML toolchains required for real-time performance.
We have been building augmented reality applications since ARKit's first public release in 2017. When Apple announced visionOS, we had a working spatial prototype within weeks of the SDK becoming available. That head start matters: spatial computing has a steep learning curve, and the frameworks are still maturing rapidly. Working with a team that has already encountered and solved the platform's edge cases means fewer surprises, faster iteration, and a product that ships on time.
Our engineering goes deeper than the high-level frameworks. We write custom Metal shaders for rendering effects that RealityKit cannot express out of the box. We build Core ML pipelines that run inference on Apple Neural Engine at 60fps without thermal throttling. We optimise USDZ assets so they render at native resolution on every supported device, from iPhone SE to Apple Vision Pro. This GPU-level expertise is what separates an AR demo from a production AR product that millions of people actually use.
Every project is handled directly by senior engineers. No account managers, no outsourcing, no handoffs. You work with the same people who architect the system, write the shaders, train the models, and debug the frame drops. That is how we have delivered AR products to Fortune 500 companies and consumer apps with millions of downloads, consistently, since 2017.
Tell us what you are building. We will outline the technical approach, timeline, and what it takes to ship.
Get in touch →