digital_ink_recognition_mlkit 0.1.1

A Flutter plugin to use Google's ML Kit Digital Ink Recognition to recognize handwritten text on a digital surface in hundreds of languages, as well as classify sketches.

Digital Ink Recognition ML Kit for Flutter

A lightweight, standalone Flutter plugin for Google ML Kit Digital Ink Recognition. It recognizes handwriting, symbols, and sketches directly on Android and iOS devices.

This package separates Digital Ink Recognition from the google_ml_kit umbrella package. It keeps only the code and native SDKs required for digital ink recognition, so your app does not need unrelated Flutter ML Kit plugins or dependencies.

Important: Android and iOS are supported. Web is not supported. This plugin is not sponsored or maintained by Google.

Installation

flutter pub add digital_ink_recognition_mlkit

Android

Configure your app with:

android {
    compileSdk = 36

    defaultConfig {
        minSdk = 24
    }
}

iOS

Requirements:

• iOS 15.5 or later
• Xcode 13.2.1 or later
• Swift 5

Set the deployment target in ios/Podfile:

platform :ios, '15.5'

Then install the Pods:

cd ios
pod install
cd ..

If the build reports a 32-bit architecture error, exclude armv7 in: Runner > Build Settings > Excluded Architectures.

Apple Silicon + iOS Simulator 26 or later

Required for this setup: Google ML Kit pods do not currently provide a native arm64-iphonesimulator slice. Add the compatibility helper below when running an iOS 26+ simulator on an Apple Silicon Mac.

1. In ios/Podfile, add this immediately after Flutter's podhelper is loaded:

require File.expand_path(
  '.symlinks/plugins/digital_ink_recognition_mlkit/ios/scripts/apple_silicon_simulator',
  __dir__,
)

2. Add this call at the end of the existing post_install block:

post_install do |installer|
  # Keep your existing post_install configuration here.

  digital_ink_mlkit_apple_silicon_simulator_patch(installer)
end

3. Reinstall the Pods:

cd ios
pod install
cd ..

The helper patches the ML Kit binary for the active build target. The same Pods installation can still be used for both Apple Silicon simulators and physical devices. It does not add another Flutter or ML Kit plugin dependency.

Quick start

import 'package:digital_ink_recognition_mlkit/digital_ink_recognition_mlkit.dart';

final recognizer = DigitalInkRecognizer(languageCode: 'en');

// The language model must be available before recognition.
final ready = await recognizer.ensureModel();
if (!ready) {
  throw Exception('Could not download the English model');
}

final ink = Ink()
  ..strokes = [
    Stroke()
      ..points = [
        StrokePoint(
          x: 20,
          y: 30,
          t: DateTime.now().millisecondsSinceEpoch,
        ),
        StrokePoint(
          x: 35,
          y: 45,
          t: DateTime.now().millisecondsSinceEpoch,
        ),
      ],
  ];

final candidates = await recognizer.recognize(ink);
if (candidates.isNotEmpty) {
  print(candidates.first.text);
}

await recognizer.close();

Each Stroke represents one continuous pen or finger movement. Each StrokePoint contains its x and y coordinates and timestamp t in milliseconds.

Language models use BCP-47 tags such as en, vi, ja, and zh-Hani. See the supported model list.

Model management:

await recognizer.isModelDownloaded();
await recognizer.downLoadModel();
await recognizer.deleteModel();

Provide the writing area and preceding text to improve recognition accuracy:

final context = DigitalInkRecognitionContext(
  preContext: 'Hello',
  writingArea: WritingArea(width: 320, height: 240),
);

final candidates = await recognizer.recognize(ink, context: context);

See the complete app in example.

How it works and upstream sources

The plugin uses Flutter Platform Channels to call the native ML Kit SDKs:

• Android: com.google.mlkit:digital-ink-recognition:19.0.0
• iOS: GoogleMLKit/DigitalInkRecognition ~> 9.0.0

The API and plugin structure are based on Google's Digital Ink Recognition documentation and the open-source google_mlkit_digital_ink_recognition implementation.

Dart sends stroke data through a platform channel and receives recognition candidates. Recognition itself runs in Google's native SDK on the device. Internet access is only required when downloading a language model for the first time.

License

This project is available under the MIT License. Anyone may use, copy, modify, distribute, or use the source commercially. No author credit or copyright notice is required in the application's user interface. Keep the LICENSE file when redistributing the software or a substantial portion of its source code.

Google ML Kit and other third-party libraries remain subject to their own licenses and terms.