SurfaceRecorderSDK

Record anything on iOS to video — an AR scene, the camera, or any view — with synced audio, without ReplayKit's permission prompt.

SurfaceRecorderSDK is a small, Swift-concurrency-native recording SDK. You give it a source (RealityKit AR, the device camera, or a UIKit/SwiftUI view) and it writes an MP4 — frames recycled from a pool, audio time-synced to the video clock, the whole pipeline driven by a single actor.

(GIF placeholder: recording an AR scene with a live FPS / CPU / MEM HUD)

import SurfaceRecorderSDK

let recorder = SurfaceRecorder(
    source: ARViewFrameSource(arView),   // or CameraFrameSource(), UIViewFrameSource(window), …
    audio: .microphone                   // or .none (default — no permission prompt)
)

try await recorder.start()
// … user does stuff …
let url = try await recorder.stop()      // MP4 in the temp dir, ready to share or save

Features

• 🎥 Four sources, one recorder — AR (camera + 3D), AR camera-only, the device camera, or any view.
• 🚫 No ReplayKit prompt — in-app capture of your own content; nothing to authorize.
• ⚡ GPU-fast AR — ARViewFrameSource taps postProcess and never touches the CPU rasterizer.
• 🎙 Audio that lines up — microphone capture with CMSampleBuffer time-correction and AVAudioSession/ARSession coexistence. Off by default — no mic prompt unless you ask.
• 📐 Resolution cap — downscale large screens to keep capture cheap.
• 📊 Live perf readout — SurfacePerformanceMonitor (FPS / CPU / memory) you can show on screen.
• 🧱 Swift 6, async-native — actors, AsyncSequence, strict concurrency.

Install

Swift Package Manager:

.package(url: "https://github.com/NazarKozak/SurfaceRecorderSDK.git", from: "0.1.0")

…and add "SurfaceRecorderSDK" to your target's dependencies. Requires iOS 17+.

Tracking the latest instead of a release? Use branch: "main".

Sources

Everything records through a FrameSource. Pick the one that matches what you want on tape:

Source	Captures	How	Cost
ARViewFrameSource	RealityKit AR: camera + 3D content	ARView.renderCallbacks.postProcess → Metal blit	🟢 GPU
RealityKitFrameSource	AR camera feed only (no 3D/UI)	zero-copy ARFrame buffer, GPU rotate/convert	🟢 GPU
CameraFrameSource	a plain device camera (back/front), with preview	AVCaptureSession video output	🟢 GPU
UIViewFrameSource	any UIView / the whole screen (UIKit + SwiftUI)	single-pass drawHierarchy, pooled buffers	🟡 CPU

// AR scene — camera + 3D, recommended for ARKit/RealityKit.
SurfaceRecorder(source: ARViewFrameSource(arView), audio: .microphone)

// AR camera feed only (clean), with rotation control.
SurfaceRecorder(source: RealityKitFrameSource(arView, orientation: .portrait))

// The device camera (like a camera app). Expose `.session` to show a preview.
let camera = CameraFrameSource(position: .back, fps: 60)
camera.startRunning()                       // drives the preview
SurfaceRecorder(source: camera)

// Any view or the key window — screen capture, NO ReplayKit prompt.
SurfaceRecorder(source: UIViewFrameSource(window, maxDimension: 1080))

No-permission screen capture

UIViewFrameSource renders your app's own view hierarchy with drawHierarchy, so it never triggers the system screen-recording prompt. Trade-off vs ReplayKit: it captures only your app (not other apps or system UI) — exactly what an in-app recorder wants.

drawHierarchy runs on the main thread and its cost scales with output pixels, so use maxDimension: to cap resolution on large screens. To also capture Metal-backed content (an ARView) in the same pass, pass afterScreenUpdates: true (heavier).

Audio

Audio is opt-in — the recorder defaults to .none, so nothing prompts the user:

SurfaceRecorder(source: …, audio: .microphone)   // mic, time-synced to the video clock
SurfaceRecorder(source: …, audio: .none)          // silent (default)

The microphone path manages AVAudioSession (interruptions, route changes) and re-times each sample buffer to the video session clock, so audio and video stay in lockstep even alongside ARKit.

Output & sharing

stop() returns the MP4 URL in the temp directory. Move it where you like, or hand it straight to a ShareLink:

let url = try await recorder.stop()
ShareLink(item: url)                // AirDrop / Save Video / Files

Performance

SurfaceRecorderSDK is built to be cheap:

• GPU AR path — ARViewFrameSource blits the composited render texture into a pixel buffer; no CPU rasterization.
• Zero-copy AR camera — RealityKitFrameSource hands the ARFrame buffer straight to a GPU convert/rotate.
• Single-pass UI capture — UIViewFrameSource renders into the pixel buffer's backing CGContext, no intermediate UIImage.
• Recycled buffers — every source pulls from a CVPixelBufferPool.
• Serialized, lock-free — one actor owns the writer; no manual locking.

Watch it live with SurfacePerformanceMonitor:

@State private var perf = SurfacePerformanceMonitor()
// …
.onAppear { perf.start() }
Text(perf.summary)   // "60 fps · 8% cpu · 124 MB"

Demo

Open Demo/SurfaceRecorderSDKDemo.xcodeproj in Xcode, pick a simulator or your device, and run. The demo has:

• three tabs — Screen, ARKit, Camera;
• an FPS selector (30 / 60 / Max) and a microphone toggle;
• a gallery that saves recordings to Documents and plays them back;
• a live FPS / CPU / MEM overlay.

AR and camera capture need a real device (the simulator has no camera).

The project depends on the local package and uses Xcode's file-system-synchronized groups — drop a .swift file into Demo/Sources/ and it's picked up automatically. (swift build only compiles the library, not the iOS demo.)

Why SurfaceRecorderSDK

	ReplayKit	ARVideoKit	SurfaceRecorderSDK
No permission prompt (in-app capture)	❌	✅	✅
RealityKit ARView (camera + 3D)	⚠️	❌	✅
Plain device camera + preview	—	—	✅
Any UIKit/SwiftUI view / screen	❌	❌	✅
Audio time-synced to the video clock	✅	⚠️	✅
Swift 6 / async-native	❌	❌	✅
Maintained for the RealityKit era	✅	⚠️	✅

Requirements

• iOS 17+
• Swift 6 / Xcode 16+
• For AR/camera capture: a real device. Add NSCameraUsageDescription (and NSMicrophoneUsageDescription if you record audio) to your Info.plist.

Roadmap

• GPU AR recording — camera + 3D (ARViewFrameSource)
• AR camera-only, zero-copy with orientation control (RealityKitFrameSource)
• Device camera with preview (CameraFrameSource)
• No-permission screen/UI capture, single-pass + resolution cap (UIViewFrameSource)
• Mic audio with time-sync + AVAudioSession/ARKit coexistence
• Live performance monitor
• GPU overlay compositing — UI/HUD over AR at full frame rate (ARViewFrameSource.setOverlay)
• GPU overlay for CameraFrameSource too
• ARSCNViewFrameSource (SceneKit) + "migrate from ARVideoKit" guide
• Photo / GIF / Live Photo capture
• visionOS, HEVC, pause/resume

License

MIT — see LICENSE.