live_beauty_filter

A Flutter plugin for iOS that applies a real-time milky/soft beauty filter to the live camera feed, entirely on the GPU. Built on AVFoundation + CoreImage + Metal — zero CPU pixel processing.

Features

  • Live camera preview with milky filter applied in real-time
  • GPU-only pipeline — no frame drops, no battery drain from CPU processing
  • Adjustable filter intensity at runtime (0.0 → 1.0)
  • Front camera with correct mirroring and portrait orientation
  • Zero-copy frame delivery to Flutter via Texture widget
  • Filter chain: Gaussian blur → Bloom/Glow → Color grade (brightness + contrast lift)

Platform support

Platform Supported
iOS ✅ iOS 14+
Android ❌ Not supported

Installation

This is a local plugin — add it to your pubspec.yaml:

dependencies:
  live_beauty_filter:
    path: ../live_beauty_filter

Add camera permission to your app's ios/Runner/Info.plist:

<key>NSCameraUsageDescription</key>
<string>Camera is used for the live beauty filter preview.</string>

Plugin registration is automatic via Flutter's GeneratedPluginRegistrant — no AppDelegate.swift changes needed.

Usage

Basic live preview

import 'package:live_beauty_filter/live_beauty_filter.dart';

class CameraScreen extends StatefulWidget {
  const CameraScreen({super.key});

  @override
  State<CameraScreen> createState() => _CameraScreenState();
}

class _CameraScreenState extends State<CameraScreen> {
  final _controller = MilkyFilterController();

  @override
  void initState() {
    super.initState();
    _controller.initialize().then((_) => setState(() {}));
  }

  @override
  void dispose() {
    _controller.dispose();
    super.dispose();
  }

  @override
  Widget build(BuildContext context) {
    if (!_controller.isInitialized) {
      return const Center(child: CircularProgressIndicator());
    }
    return Texture(textureId: _controller.textureId!);
  }
}

Adjusting filter intensity

// intensity: 0.0 = no filter, 1.0 = full milky
await _controller.setFilterIntensity(0.7);

Drop-in widget with built-in intensity slider

MilkyCameraView()

How the filter works

The milky look is achieved by chaining three CoreImage GPU passes on every frame:

CVPixelBuffer (raw frame)
    │
    ▼
CIGaussianBlur          — light softening (radius scales with intensity)
    │
    ▼
Bloom / Screen blend    — blurred copy composited over original for glow
    │
    ▼
CIColorControls         — brightness +, contrast -, saturation slight -
    │
    ▼
CIColorPolynomial       — raises black point (lifted shadows, milky whites)
    │
    ▼
CVPixelBuffer (filtered) → FlutterTexture → Texture() widget

All operations run on the Metal GPU via a CIContext created with useSoftwareRenderer: false. Frames are pooled with CVPixelBufferPool to eliminate per-frame heap allocation.

API reference

MilkyFilterController

Method Description
initialize() Starts the camera session and registers the GPU texture
setFilterIntensity(double) Updates filter strength live (0.0–1.0, default 0.7)
dispose() Stops capture and releases the texture
textureId The Flutter texture ID — pass to Texture(textureId: ...)
isInitialized Whether the camera and filter pipeline are ready

MilkyFilterPipeline (iOS internal)

Property Type Description
intensity Float Master filter strength, updated from Flutter via MethodChannel

Tuning the filter

Edit the constants in MilkyFilterPipeline.swift to adjust the look:

// MilkyFilterPipeline.swift

// Blur softness (max radius at intensity 1.0)
let blurRadius = Double(intensity) * 4.0          // increase for dreamier blur

// Bloom strength (how much glow blends back in)
let bloomStrength = Double(intensity) * 0.45      // increase for stronger glow

// Color grade
let brightnessBoost      = Double(intensity) * 0.08   // raise for brighter
let contrastReduction    = 1.0 - Double(intensity) * 0.15  // lower = flatter/milkier
let saturationReduction  = 1.0 - Double(intensity) * 0.12  // lower = more faded

// Blue channel lift in CIColorPolynomial
// inputBlueCoefficients: CIVector(x: intensity * 0.06, ...)
// Increase x for a cooler (more blue-white) milky tone
// Decrease for a warmer tone

Testing

Integration tests must run on a physical device — the camera does not work on Simulator.

Grant camera permission by running the example app once manually first:

cd example
flutter run -d <your-device-id>

Then run the integration tests:

flutter test integration_test/plugin_integration_test.dart -d <your-device-id>

Expected output:

+2: All tests passed!

Requirements

  • iOS 14.0+
  • Xcode 14+
  • A physical device (camera does not work on Simulator)
  • Swift 5.7+

File structure

live_beauty_filter/
├── ios/
│   ├── Classes/
│   │   ├── LiveBeautyFilterPlugin.swift      # Flutter plugin entry + MethodChannel
│   │   ├── MilkyCameraController.swift       # AVCapture session + frame delivery
│   │   └── MilkyFilterPipeline.swift         # CIFilter GPU chain (blur + bloom + grade)
│   ├── Resources/
│   │   └── PrivacyInfo.xcprivacy             # App Store privacy manifest
│   └── live_beauty_filter.podspec
├── lib/
│   ├── live_beauty_filter.dart               # Barrel export (main entry point)
│   ├── milky_filter_controller.dart          # MethodChannel Dart wrapper
│   ├── milky_camera_view.dart                # Drop-in Flutter widget with slider
│   ├── live_beauty_filter_platform_interface.dart
│   └── live_beauty_filter_method_channel.dart
├── example/
│   ├── lib/
│   │   └── main.dart                         # Example app — fullscreen filter preview
│   ├── integration_test/
│   │   └── plugin_integration_test.dart      # Integration tests (run on physical device)
│   └── ios/
│       └── Runner/
│           └── Info.plist                    # Must contain NSCameraUsageDescription
├── pubspec.yaml
└── README.md

License

This project is licensed under the MIT License see the LICENSE file for details.

Libraries

live_beauty_filter
live_beauty_filter_method_channel
live_beauty_filter_platform_interface
milky_camera_view
milky_filter_controller