✨ deepar_kit

Real-time AR filters for your live streams — DeepAR meets Agora, in one line.

Pipe DeepAR AR-processed frames straight into Agora RTC as an external video source — or use DeepAR standalone as a pure camera filter.

🚀 Why deepar_kit?

Integrating augmented-reality beauty and face filters into a real-time video call usually means wiring raw camera buffers, format conversions, and low-level frame pushes by hand. deepar_kit collapses all of that into a single, resilient bridge — and works just as well as a standalone camera filter with no streaming at all.

🔗 Automatic frame forwarding — DeepAR output flows into Agora with zero manual plumbing
🎭 First-class effects — load .deepar filters and switch cameras through convenience methods
🧩 Works with or without Agora — drop the bridge and DeepAR runs as a standalone camera filter
🛡️ Error-resilient — dropped frames never crash your stream
📊 Observable — built-in frame-count tracking for diagnostics
🧹 Clean lifecycle — deterministic start / stop / dispose

📦 Installation

Add the package to your pubspec.yaml:

dependencies:
  deepar_kit: ^0.1.0
  # Automatically pulls in flutter_deepar and agora_rtc_engine

Then run:

flutter pub get

🎯 Two ways to use it

One package, two entry points — import only what you need:

Use case	Import	What you get
🎥 Camera filter only (no Agora)	`package:deepar_kit/deepar_filter.dart`	`DeepARController` + effects. No bridge, no Agora setup.
📡 Filters + live streaming	`package:deepar_kit/deepar_kit.dart`	Everything above plus `DeepARAgoraBridge`.

ℹ️ Both live in the same package, so agora_rtc_engine is present in the dependency tree either way — but with the filter-only import you never touch a single Agora type or API.

🔐 Separation of concerns

Your app stays the sole owner of Agora. This package never asks for, reads, or stores your Agora App ID, token, or channel — you create and initialize the RtcEngine yourself and simply hand it to the bridge. DeepAR and Agora remain fully decoupled:

Want AR filters in a live stream? → attach the bridge.
Want AR filters as a plain camera filter? → use DeepARController alone, no Agora involved.

⚡ Quick Start

import 'package:deepar_kit/deepar_kit.dart';
import 'package:agora_rtc_engine/agora_rtc_engine.dart';

// 1️⃣  Initialize DeepAR
final deepar = DeepARController();
await deepar.initialize(licenseKey: 'YOUR_DEEPAR_KEY');
await deepar.startCapture();

// 2️⃣  Your app owns Agora — App ID/token never enter this package
final agoraEngine = createAgoraRtcEngine();
await agoraEngine.initialize(RtcEngineContext(appId: 'YOUR_AGORA_APP_ID'));

// 3️⃣  Bridge them and go live — one call does all the Agora wiring
final bridge = DeepARAgoraBridge(deepAR: deepar, agoraEngine: agoraEngine);
await bridge.enable();
// 🎉 AR-processed frames now stream into Agora automatically!

// 4️⃣  Load an effect at any time
await bridge.loadEffect('effects/my_filter.deepar');

// 5️⃣  Tear down cleanly
await bridge.disable();
await deepar.stopCapture();
bridge.dispose();

🎥 Standalone Mode (no Agora)

Don't need streaming? Import the filter-only entry point — no bridge, no Agora API, ever.

import 'package:deepar_kit/deepar_filter.dart';

final deepar = DeepARController();
await deepar.initialize(licenseKey: 'YOUR_DEEPAR_KEY');
await deepar.startCapture();
await deepar.loadEffect('effects/my_filter.deepar');
// Render deepar's preview widget directly in your UI.

🔬 How It Works

The bridge subscribes to DeepARController.frameStream and pushes each processed frame to Agora via MediaEngine.pushVideoFrame(). Under the hood it handles:

Concern	Handled
Pixel format	🟢 RGBA (Android) & BGRA (iOS) auto-detection
Timing	🟢 Per-frame timestamping
Resilience	🟢 Frame-push failures are non-fatal
Lifecycle	🟢 Deterministic start / stop / dispose

📋 Requirements

Flutter >= 3.10.0
Dart SDK >= 3.0.0
A DeepAR license key
An Agora App ID

🤝 Contributing

Contributions, issues, and feature requests are welcome! Feel free to open an issue or submit a pull request.

👤 Author

Zaman Sheikh

GitHub: @zamansheikh

📄 License

If this package helped you, consider giving it a ⭐ on GitHub!