flutter_native_vision_camera 0.0.1

High-performance Flutter FFI camera plugin with zero-copy preview and real-time frame processing.

Flutter Native Vision Camera

A high-performance, FFI-powered camera plugin for Flutter that puts efficiency and hardware control first.

Built for developers who need more than just a preview: Real-time AI, low-latency filters, and professional-grade video control.

Why use this instead of camera?

The official camera package is great for simple use cases, but it often falls short when building high-performance vision applications (AI, QR scanning, custom filters). This package was built to solve those limitations.

🚀 Key Advantages

Feature	camera (standard)	flutter_native_vision_camera
Rendering	Platform Channels (YUV -> Bitmap -> Skia)	Zero-Copy GPU Textures (OES -> Flutter Texture)
Orientation	Matches UI Orientation (Buggy for video)	Hardware-Level Sensing (Correct even if UI is locked)
Frame Processing	Async via Isolate (High Latency)	Synchronous Background FFI (Near Zero Latency)
QR/Barcodes	Third-party plugin required	Optimized MLKit Integration built-in
Zoom/Focus	Basic level support	Unified request management (Works during recording)

Features

• Zero-Copy Rendering: Frames go directly from the camera sensor to the GPU and then to Flutter's Texture widget. No expensive copying or CPU-side bitmap conversion.
• Physical Orientation Engine: High-fidelity rotation sensing using native OrientationEventListener. Recorded videos and previews always have the correct orientation, even on orientation-locked apps.
• Integrated MLKit: High-speed, hardware-accelerated barcode and QR code scanning.
• Unified Request Loop: Change zoom, torch, exposure, or recording state without dropping frame buffers.
• FFI Background Pipeline: Process camera frames in C++/Rust/Dart FFI with zero memory copying.

Frame Processors

Frame processors allow you to run code for every frame the camera captures. Unlike the standard camera plugin, flutter_native_vision_camera runs these synchronously on a background thread.

How it works

1. Zero Latency: Frames are delivered to the processor as soon as they are available from the sensor.
2. Background Execution: Your code runs on a separate background Isolate or in Native C++, so the main UI thread stays at a silky smooth 60/120 FPS.
3. Memory Safety: The Frame object is valid only for the duration of the callback. If you need to keep data, copy it out of the frame's buffer.

Dart Frame Processor

Use for ML tasks (using google_mlkit_* or tflite_flutter) or image manipulation in Dart.

await controller.setFrameProcessor((frame) {
  // 'frame' contains pointers to YUV/RGB buffers
  final bytes = frame.getPlane(0).bytes; // Direct access to Y-plane
  
  // Do your analysis here...
});

🚀 High-Performance FFI Plugins

One of the unique features of flutter_native_vision_camera is the ability to write Synchronous C++ Plugins.

Instead of sending expensive image data back and forth over Method Channels or into separate Dart Isolates, you can hook directly into the native frame loop.

// Your high-performance C++ code
void onFrame(FrameHandle frame) {
    void* y_plane = Frame_getPlanePointer(frame, 0);
    // Do heavy AI math here on the background thread
}

Getting Started

Installation

Add to your pubspec.yaml:

dependencies:
  flutter_native_vision_camera: ^1.0.0

Basic Usage

final controller = CameraController();

// 1. Initialize with a device
await controller.initialize(
  devices.first,
  enableVideo: true,
  codeScanner: CodeScannerConfiguration(),
);

// 2. Start streaming
await controller.setActive(true);

// 3. Listen for codes
controller.onCodeScanned.listen((codes) {
  print('Detected: ${codes.first.value}');
});

// 4. Render in your widget tree
ValueListenableBuilder<CameraState>(
  valueListenable: controller,
  builder: (context, state, _) {
    if (state == CameraState.uninitialized) return CircularProgressIndicator();
    return Texture(textureId: controller.textureId!);
  },
)

Lifecycle Management

Keeping the camera hardware active is resource-intensive. You must dispose of the controller when it's no longer needed to release the hardware and stop background threads.

@override
void dispose() {
  // Shuts down the camera, sessions, and background isolates
  controller.dispose(); 
  super.dispose();
}

Video Recording & Zooming

Unlike many other plugins, you can smoothly zoom or toggle the torch while recording video without causing any frame drops or freezes.

await controller.startRecording('/path/to/video.mp4');

// This call seamlessly updates the ongoing recording request
await controller.setZoom(2.5); 

await controller.stopRecording();

Platform Support

Platform	Support	Notes
Android	✅ Full	Camera2, MLKit, FFI, Recording
iOS	⚠️ Partial	AVFoundation, Vision API, Preview only. Recording & FFI coming soon.

Credits & Attribution

This package is a Flutter port and expansion of the concepts introduced by react-native-vision-camera, originally created by Marc Rousavy.

We aim to bring the same high-performance, low-level camera control to the Flutter ecosystem, while leveraging Flutter's unique strengths like synchronous FFI and specialized Texture rendering.

License

MIT