Add TTS API and examples for Dart (#1010)

.github/scripts/test-dart.sh

@@ -4,6 +4,22 @@ set -ex
 
 cd dart-api-examples
 
+pushd tts
+
+echo '----------piper tts----------'
+./run-piper.sh
+rm -rf vits-piper-*
+
+echo '----------coqui tts----------'
+./run-coqui.sh
+rm -rf vits-coqui-*
+
+echo '----------zh tts----------'
+./run-zh.sh
+rm -rf sherpa-onnx-*
+
+popd # tts
+
 pushd streaming-asr
 
 echo '----------streaming zipformer ctc HLG----------'

.github/workflows/test-dart.yaml

@@ -92,5 +92,6 @@ jobs:
           cp scripts/dart/vad-pubspec.yaml dart-api-examples/vad/pubspec.yaml
           cp scripts/dart/non-streaming-asr-pubspec.yaml dart-api-examples/non-streaming-asr/pubspec.yaml
           cp scripts/dart/streaming-asr-pubspec.yaml dart-api-examples/streaming-asr/pubspec.yaml
+          cp scripts/dart/tts-pubspec.yaml dart-api-examples/tts/pubspec.yaml
 
           .github/scripts/test-dart.sh

dart-api-examples/streaming-asr/README.md

@@ -1,6 +1,7 @@
 # Introduction
 
 This folder contains examples for streaming ASR with Dart API.
+
 | File | Description|
 |------|------------|
 |[./bin/nemo-transducer.dart](./bin/nemo-transducer.dart)| Use a NeMo transducer model for speech recognition. See [./run-nemo-transducer.sh](./run-nemo-transducer.sh)|

dart-api-examples/tts/.gitignore

@@ -0,0 +1,3 @@
+# https://dart.dev/guides/libraries/private-files
+# Created by `dart pub`
+.dart_tool/

dart-api-examples/tts/CHANGELOG.md

@@ -0,0 +1,3 @@
+## 1.0.0
+
+- Initial version.

dart-api-examples/tts/README.md

@@ -0,0 +1,10 @@
+# Introduction
+
+This folder contains examples for text to speech with Dart API.
+
+| File | Description|
+|------|------------|
+|[./bin/piper.dart](./bin/piper.dart)| Use a Piper tts model for text to speech. See [./run-piper.sh](./run-piper.sh)|
+|[./bin/coqui.dart](./bin/coqui.dart)| Use a Coqui tts model for text to speech. See [./run-coqui.sh](./run-coqui.sh)|
+|[./bin/zh.dart](./bin/zh.dart)| Use a Chinese VITS tts model for text to speech. See [./run-zh.sh](./run-zh.sh)|
+

dart-api-examples/tts/analysis_options.yaml

@@ -0,0 +1,30 @@
+# This file configures the static analysis results for your project (errors,
+# warnings, and lints).
+#
+# This enables the 'recommended' set of lints from `package:lints`.
+# This set helps identify many issues that may lead to problems when running
+# or consuming Dart code, and enforces writing Dart using a single, idiomatic
+# style and format.
+#
+# If you want a smaller set of lints you can change this to specify
+# 'package:lints/core.yaml'. These are just the most critical lints
+# (the recommended set includes the core lints).
+# The core lints are also what is used by pub.dev for scoring packages.
+
+include: package:lints/recommended.yaml
+
+# Uncomment the following section to specify additional rules.
+
+# linter:
+#   rules:
+#     - camel_case_types
+
+# analyzer:
+#   exclude:
+#     - path/to/excluded/files/**
+
+# For more information about the core and recommended set of lints, see
+# https://dart.dev/go/core-lints
+
+# For additional information about configuring this file, see
+# https://dart.dev/guides/language/analysis-options

dart-api-examples/tts/bin/coqui.dart

@@ -0,0 +1,69 @@
+// Copyright (c)  2024  Xiaomi Corporation
+import 'dart:io';
+import 'dart:typed_data';
+
+import 'package:args/args.dart';
+import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
+
+import './init.dart';
+
+void main(List<String> arguments) async {
+  await initSherpaOnnx();
+
+  final parser = ArgParser()
+    ..addOption('model', help: 'Path to the ONNX model')
+    ..addOption('tokens', help: 'Path to tokens.txt')
+    ..addOption('text', help: 'Text to generate TTS for')
+    ..addOption('output-wav', help: 'Filename to save the generated audio')
+    ..addOption('speed', help: 'Speech speed', defaultsTo: '1.0')
+    ..addOption(
+      'sid',
+      help: 'Speaker ID to select. Used only for multi-speaker TTS',
+      defaultsTo: '0',
+    );
+  final res = parser.parse(arguments);
+  if (res['model'] == null ||
+      res['tokens'] == null ||
+      res['output-wav'] == null ||
+      res['text'] == null) {
+    print(parser.usage);
+    exit(1);
+  }
+  final model = res['model'] as String;
+  final tokens = res['tokens'] as String;
+  final text = res['text'] as String;
+  final outputWav = res['output-wav'] as String;
+  var speed = double.tryParse(res['speed'] as String) ?? 1.0;
+  final sid = int.tryParse(res['sid'] as String) ?? 0;
+
+  if (speed == 0) {
+    speed = 1.0;
+  }
+
+  final vits = sherpa_onnx.OfflineTtsVitsModelConfig(
+    model: model,
+    tokens: tokens,
+    lengthScale: 1 / speed,
+  );
+
+  final modelConfig = sherpa_onnx.OfflineTtsModelConfig(
+    vits: vits,
+    numThreads: 1,
+    debug: true,
+  );
+  final config = sherpa_onnx.OfflineTtsConfig(
+    model: modelConfig,
+    maxNumSenetences: 1,
+  );
+
+  final tts = sherpa_onnx.OfflineTts(config);
+  final audio = tts.generate(text: text, sid: sid, speed: speed);
+  tts.free();
+
+  sherpa_onnx.writeWave(
+    filename: outputWav,
+    samples: audio.samples,
+    sampleRate: audio.sampleRate,
+  );
+  print('Saved to ${outputWav}');
+}

dart-api-examples/tts/bin/init.dart

@@ -0,0 +1 @@
+../../vad/bin/init.dart

dart-api-examples/tts/bin/piper.dart

@@ -0,0 +1,80 @@
+// Copyright (c)  2024  Xiaomi Corporation
+import 'dart:io';
+import 'dart:typed_data';
+
+import 'package:args/args.dart';
+import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
+
+import './init.dart';
+
+void main(List<String> arguments) async {
+  await initSherpaOnnx();
+
+  final parser = ArgParser()
+    ..addOption('model', help: 'Path to the ONNX model')
+    ..addOption('tokens', help: 'Path to tokens.txt')
+    ..addOption('data-dir', help: 'Path to espeak-ng-data directory')
+    ..addOption('text', help: 'Text to generate TTS for')
+    ..addOption('output-wav', help: 'Filename to save the generated audio')
+    ..addOption('speed', help: 'Speech speed', defaultsTo: '1.0')
+    ..addOption(
+      'sid',
+      help: 'Speaker ID to select. Used only for multi-speaker TTS',
+      defaultsTo: '0',
+    );
+  final res = parser.parse(arguments);
+  if (res['model'] == null ||
+      res['tokens'] == null ||
+      res['data-dir'] == null ||
+      res['output-wav'] == null ||
+      res['text'] == null) {
+    print(parser.usage);
+    exit(1);
+  }
+  final model = res['model'] as String;
+  final tokens = res['tokens'] as String;
+  final dataDir = res['data-dir'] as String;
+  final text = res['text'] as String;
+  final outputWav = res['output-wav'] as String;
+  var speed = double.tryParse(res['speed'] as String) ?? 1.0;
+  final sid = int.tryParse(res['sid'] as String) ?? 0;
+
+  if (speed == 0) {
+    speed = 1.0;
+  }
+
+  final vits = sherpa_onnx.OfflineTtsVitsModelConfig(
+    model: model,
+    tokens: tokens,
+    dataDir: dataDir,
+    lengthScale: 1 / speed,
+  );
+
+  final modelConfig = sherpa_onnx.OfflineTtsModelConfig(
+    vits: vits,
+    numThreads: 1,
+    debug: true,
+  );
+  final config = sherpa_onnx.OfflineTtsConfig(
+    model: modelConfig,
+    maxNumSenetences: 1,
+  );
+
+  final tts = sherpa_onnx.OfflineTts(config);
+  final audio = tts.generateWithCallback(
+      text: text,
+      sid: sid,
+      speed: speed,
+      callback: (Float32List samples) {
+        print('${samples.length} samples received');
+        // You can play samples in a separate thread/isolate
+      });
+  tts.free();
+
+  sherpa_onnx.writeWave(
+    filename: outputWav,
+    samples: audio.samples,
+    sampleRate: audio.sampleRate,
+  );
+  print('Saved to ${outputWav}');
+}

dart-api-examples/tts/bin/zh.dart

@@ -0,0 +1,86 @@
+// Copyright (c)  2024  Xiaomi Corporation
+import 'dart:io';
+import 'dart:typed_data';
+
+import 'package:args/args.dart';
+import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
+
+import './init.dart';
+
+void main(List<String> arguments) async {
+  await initSherpaOnnx();
+
+  final parser = ArgParser()
+    ..addOption('model', help: 'Path to the ONNX model')
+    ..addOption('tokens', help: 'Path to tokens.txt')
+    ..addOption('lexicon', help: 'Path to lexicon.txt')
+    ..addOption(
+      'dict-dir',
+      help: 'Path to jieba dict directory',
+      defaultsTo: '',
+    )
+    ..addOption('rule-fsts', help: 'Path to rule fsts', defaultsTo: '')
+    ..addOption('rule-fars', help: 'Path to rule fars', defaultsTo: '')
+    ..addOption('text', help: 'Text to generate TTS for')
+    ..addOption('output-wav', help: 'Filename to save the generated audio')
+    ..addOption('speed', help: 'Speech speed', defaultsTo: '1.0')
+    ..addOption(
+      'sid',
+      help: 'Speaker ID to select. Used only for multi-speaker TTS',
+      defaultsTo: '0',
+    );
+  final res = parser.parse(arguments);
+  if (res['model'] == null ||
+      res['lexicon'] == null ||
+      res['tokens'] == null ||
+      res['output-wav'] == null ||
+      res['text'] == null) {
+    print(parser.usage);
+    exit(1);
+  }
+  final model = res['model'] as String;
+  final lexicon = res['lexicon'] as String;
+  final tokens = res['tokens'] as String;
+  final dictDir = res['dict-dir'] as String;
+  final ruleFsts = res['rule-fsts'] as String;
+  final ruleFars = res['rule-fars'] as String;
+  final text = res['text'] as String;
+  final outputWav = res['output-wav'] as String;
+  var speed = double.tryParse(res['speed'] as String) ?? 1.0;
+  final sid = int.tryParse(res['sid'] as String) ?? 0;
+
+  if (speed == 0) {
+    speed = 1.0;
+  }
+
+  final vits = sherpa_onnx.OfflineTtsVitsModelConfig(
+    model: model,
+    lexicon: lexicon,
+    tokens: tokens,
+    dictDir: dictDir,
+    lengthScale: 1 / speed,
+  );
+
+  final modelConfig = sherpa_onnx.OfflineTtsModelConfig(
+    vits: vits,
+    numThreads: 1,
+    debug: true,
+  );
+  final config = sherpa_onnx.OfflineTtsConfig(
+    model: modelConfig,
+    maxNumSenetences: 1,
+    ruleFsts: ruleFsts,
+    ruleFars: ruleFars,
+  );
+
+  final tts = sherpa_onnx.OfflineTts(config);
+  final audio = tts.generate(text: text, sid: sid, speed: speed);
+  tts.free();
+
+  sherpa_onnx.writeWave(
+    filename: outputWav,
+    samples: audio.samples,
+    sampleRate: audio.sampleRate,
+  );
+  print('Saved to ${outputWav}');
+}

dart-api-examples/tts/pubspec.lock

@@ -0,0 +1,96 @@
+# Generated by pub
+# See https://dart.dev/tools/pub/glossary#lockfile
+packages:
+  args:
+    dependency: "direct main"
+    description:
+      name: args
+      sha256: "7cf60b9f0cc88203c5a190b4cd62a99feea42759a7fa695010eb5de1c0b2252a"
+      url: "https://pub.dev"
+    source: hosted
+    version: "2.5.0"
+  characters:
+    dependency: transitive
+    description:
+      name: characters
+      sha256: "04a925763edad70e8443c99234dc3328f442e811f1d8fd1a72f1c8ad0f69a605"
+      url: "https://pub.dev"
+    source: hosted
+    version: "1.3.0"
+  collection:
+    dependency: transitive
+    description:
+      name: collection
+      sha256: ee67cb0715911d28db6bf4af1026078bd6f0128b07a5f66fb2ed94ec6783c09a
+      url: "https://pub.dev"
+    source: hosted
+    version: "1.18.0"
+  ffi:
+    dependency: transitive
+    description:
+      name: ffi
+      sha256: "493f37e7df1804778ff3a53bd691d8692ddf69702cf4c1c1096a2e41b4779e21"
+      url: "https://pub.dev"
+    source: hosted
+    version: "2.1.2"
+  flutter:
+    dependency: transitive
+    description: flutter
+    source: sdk
+    version: "0.0.0"
+  lints:
+    dependency: "direct dev"
+    description:
+      name: lints
+      sha256: cbf8d4b858bb0134ef3ef87841abdf8d63bfc255c266b7bf6b39daa1085c4290
+      url: "https://pub.dev"
+    source: hosted
+    version: "3.0.0"
+  material_color_utilities:
+    dependency: transitive
+    description:
+      name: material_color_utilities
+      sha256: "0e0a020085b65b6083975e499759762399b4475f766c21668c4ecca34ea74e5a"
+      url: "https://pub.dev"
+    source: hosted
+    version: "0.8.0"
+  meta:
+    dependency: transitive
+    description:
+      name: meta
+      sha256: "7687075e408b093f36e6bbf6c91878cc0d4cd10f409506f7bc996f68220b9136"
+      url: "https://pub.dev"
+    source: hosted
+    version: "1.12.0"
+  path:
+    dependency: "direct main"
+    description:
+      name: path
+      sha256: "087ce49c3f0dc39180befefc60fdb4acd8f8620e5682fe2476afd0b3688bb4af"
+      url: "https://pub.dev"
+    source: hosted
+    version: "1.9.0"
+  sherpa_onnx:
+    dependency: "direct main"
+    description:
+      name: sherpa_onnx
+      sha256: e45894f81e7c854ca96d678bcab5303036e884a7c90e9a6c4ec04c7b1ee215a8
+      url: "https://pub.dev"
+    source: hosted
+    version: "1.9.29"
+  sky_engine:
+    dependency: transitive
+    description: flutter
+    source: sdk
+    version: "0.0.99"
+  vector_math:
+    dependency: transitive
+    description:
+      name: vector_math
+      sha256: "80b3257d1492ce4d091729e3a67a60407d227c27241d6927be0130c98e741803"
+      url: "https://pub.dev"
+    source: hosted
+    version: "2.1.4"
+sdks:
+  dart: ">=3.4.0 <4.0.0"
+  flutter: ">=3.3.0"

dart-api-examples/tts/pubspec.yaml

@@ -0,0 +1,16 @@
+name: tts
+description: A sample command-line application.
+version: 1.0.0
+# repository: https://github.com/my_org/my_repo
+
+environment:
+  sdk: ^3.4.0
+
+# Add regular dependencies here.
+dependencies:
+  sherpa_onnx: ^1.9.29
+  path: ^1.9.0
+  args: ^2.5.0
+
+dev_dependencies:
+  lints: ^3.0.0

dart-api-examples/tts/run-coqui.sh

@@ -0,0 +1,28 @@
+#!/usr/bin/env bash
+
+set -ex
+
+dart pub get
+
+
+# Please visit
+# https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models
+# to download more models
+
+if [[ ! -f ./vits-coqui-de-css10/tokens.txt ]]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-coqui-de-css10.tar.bz2
+  tar xvf vits-coqui-de-css10.tar.bz2
+  rm vits-coqui-de-css10.tar.bz2
+fi
+
+# It is a character-based TTS model, so there is no need to use a lexicon
+dart run \
+  ./bin/coqui.dart \
+  --model ./vits-coqui-de-css10/model.onnx \
+  --tokens ./vits-coqui-de-css10/tokens.txt \
+  --sid 0 \
+  --speed 0.7 \
+  --text 'Alles hat ein Ende, nur die Wurst hat zwei.' \
+  --output-wav coqui-0.wav
+
+ls -lh *.wav

dart-api-examples/tts/run-piper.sh

@@ -0,0 +1,28 @@
+#!/usr/bin/env bash
+
+set -ex
+
+dart pub get
+
+
+# Please visit
+# https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models
+# to download more models
+
+if [[ ! -f ./vits-piper-en_US-libritts_r-medium/tokens.txt ]]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-piper-en_US-libritts_r-medium.tar.bz2
+  tar xf vits-piper-en_US-libritts_r-medium.tar.bz2
+  rm vits-piper-en_US-libritts_r-medium.tar.bz2
+fi
+
+dart run \
+  ./bin/piper.dart \
+  --model ./vits-piper-en_US-libritts_r-medium/en_US-libritts_r-medium.onnx \
+  --tokens ./vits-piper-en_US-libritts_r-medium/tokens.txt \
+  --data-dir ./vits-piper-en_US-libritts_r-medium/espeak-ng-data \
+  --sid 351 \
+  --speed 1.0 \
+  --text 'How are you doing? This is a speech to text example, using next generation kaldi with piper.' \
+  --output-wav piper-351.wav
+
+ls -lh *.wav

dart-api-examples/tts/run-zh.sh

@@ -0,0 +1,41 @@
+#!/usr/bin/env bash
+
+set -ex
+
+dart pub get
+
+
+# Please visit
+# https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models
+# to download more models
+
+if [[ ! -f ./sherpa-onnx-vits-zh-ll/tokens.txt ]]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/sherpa-onnx-vits-zh-ll.tar.bz2
+  tar xvf sherpa-onnx-vits-zh-ll.tar.bz2
+  rm sherpa-onnx-vits-zh-ll.tar.bz2
+fi
+
+dart run \
+  ./bin/zh.dart \
+  --model ./sherpa-onnx-vits-zh-ll/model.onnx \
+  --lexicon ./sherpa-onnx-vits-zh-ll/lexicon.txt \
+  --tokens ./sherpa-onnx-vits-zh-ll/tokens.txt \
+  --dict-dir ./sherpa-onnx-vits-zh-ll/dict \
+  --sid 2 \
+  --speed 1.0 \
+  --text '当夜幕降临，星光点点，伴随着微风拂面，我在静谧中感受着时光的流转，思念如涟漪荡漾，梦境如画卷展开，我与自然融为一体，沉静在这片宁静的美丽之中，感受着生命的奇迹与温柔。' \
+  --output-wav zh-jieba-2.wav
+
+dart run \
+  ./bin/zh.dart \
+  --model ./sherpa-onnx-vits-zh-ll/model.onnx \
+  --lexicon ./sherpa-onnx-vits-zh-ll/lexicon.txt \
+  --tokens ./sherpa-onnx-vits-zh-ll/tokens.txt \
+  --dict-dir ./sherpa-onnx-vits-zh-ll/dict \
+  --rule-fsts "./sherpa-onnx-vits-zh-ll/phone.fst,./sherpa-onnx-vits-zh-ll/date.fst,./sherpa-onnx-vits-zh-ll/number.fst" \
+  --sid 3 \
+  --speed 1.0 \
+  --text '今天是2024年6月15号，13点23分。如果有困难，请拨打110或者18920240511。123456块钱。' \
+  --output-wav zh-jieba-3.wav
+
+ls -lh *.wav

scripts/dart/tts-pubspec.yaml

@@ -0,0 +1,17 @@
+name: tts
+description: A sample command-line application.
+version: 1.0.0
+# repository: https://github.com/my_org/my_repo
+
+environment:
+  sdk: ^3.4.0
+
+# Add regular dependencies here.
+dependencies:
+  sherpa_onnx:
+    path: ../../sherpa-onnx/flutter
+  path: ^1.9.0
+  args: ^2.5.0
+
+dev_dependencies:
+  lints: ^3.0.0

sherpa-onnx/flutter/lib/sherpa_onnx.dart

@@ -8,6 +8,7 @@ export 'src/offline_stream.dart';
 export 'src/online_recognizer.dart';
 export 'src/online_stream.dart';
 export 'src/speaker_identification.dart';
+export 'src/tts.dart';
 export 'src/vad.dart';
 export 'src/wave_reader.dart';
 export 'src/wave_writer.dart';

sherpa-onnx/flutter/lib/src/sherpa_onnx_bindings.dart

@@ -2,6 +2,55 @@
 import 'dart:ffi';
 import 'package:ffi/ffi.dart';
 
+final class SherpaOnnxOfflineTtsVitsModelConfig extends Struct {
+  external Pointer<Utf8> model;
+  external Pointer<Utf8> lexicon;
+  external Pointer<Utf8> tokens;
+  external Pointer<Utf8> dataDir;
+
+  @Float()
+  external double noiseScale;
+
+  @Float()
+  external double noiseScaleW;
+
+  @Float()
+  external double lengthScale;
+
+  external Pointer<Utf8> dictDir;
+}
+
+final class SherpaOnnxOfflineTtsModelConfig extends Struct {
+  external SherpaOnnxOfflineTtsVitsModelConfig vits;
+  @Int32()
+  external int numThreads;
+
+  @Int32()
+  external int debug;
+
+  external Pointer<Utf8> provider;
+}
+
+final class SherpaOnnxOfflineTtsConfig extends Struct {
+  external SherpaOnnxOfflineTtsModelConfig model;
+  external Pointer<Utf8> ruleFsts;
+
+  @Int32()
+  external int maxNumSenetences;
+
+  external Pointer<Utf8> ruleFars;
+}
+
+final class SherpaOnnxGeneratedAudio extends Struct {
+  external Pointer<Float> samples;
+
+  @Int32()
+  external int n;
+
+  @Int32()
+  external int sampleRate;
+}
+
 final class SherpaOnnxFeatureConfig extends Struct {
   @Int32()
   external int sampleRate;
@@ -218,6 +267,8 @@ final class SherpaOnnxSpeakerEmbeddingExtractorConfig extends Struct {
   external Pointer<Utf8> provider;
 }
 
+final class SherpaOnnxOfflineTts extends Opaque {}
+
 final class SherpaOnnxCircularBuffer extends Opaque {}
 
 final class SherpaOnnxVoiceActivityDetector extends Opaque {}
@@ -234,6 +285,60 @@ final class SherpaOnnxSpeakerEmbeddingExtractor extends Opaque {}
 
 final class SherpaOnnxSpeakerEmbeddingManager extends Opaque {}
 
+typedef SherpaOnnxCreateOfflineTtsNative = Pointer<SherpaOnnxOfflineTts>
+    Function(Pointer<SherpaOnnxOfflineTtsConfig>);
+
+typedef SherpaOnnxCreateOfflineTts = SherpaOnnxCreateOfflineTtsNative;
+
+typedef SherpaOnnxDestroyOfflineTtsNative = Void Function(
+    Pointer<SherpaOnnxOfflineTts>);
+
+typedef SherpaOnnxDestroyOfflineTts = void Function(
+    Pointer<SherpaOnnxOfflineTts>);
+
+typedef SherpaOnnxOfflineTtsSampleRateNative = Int32 Function(
+    Pointer<SherpaOnnxOfflineTts>);
+
+typedef SherpaOnnxOfflineTtsSampleRate = int Function(
+    Pointer<SherpaOnnxOfflineTts>);
+
+typedef SherpaOnnxOfflineTtsNumSpeakersNative = Int32 Function(
+    Pointer<SherpaOnnxOfflineTts>);
+
+typedef SherpaOnnxOfflineTtsNumSpeakers = int Function(
+    Pointer<SherpaOnnxOfflineTts>);
+
+typedef SherpaOnnxOfflineTtsGenerateNative = Pointer<SherpaOnnxGeneratedAudio>
+    Function(Pointer<SherpaOnnxOfflineTts>, Pointer<Utf8>, Int32, Float);
+
+typedef SherpaOnnxOfflineTtsGenerate = Pointer<SherpaOnnxGeneratedAudio>
+    Function(Pointer<SherpaOnnxOfflineTts>, Pointer<Utf8>, int, double);
+
+typedef SherpaOnnxDestroyOfflineTtsGeneratedAudioNative = Void Function(
+    Pointer<SherpaOnnxGeneratedAudio>);
+
+typedef SherpaOnnxDestroyOfflineTtsGeneratedAudio = void Function(
+    Pointer<SherpaOnnxGeneratedAudio>);
+
+typedef SherpaOnnxGeneratedAudioCallbackNative = Void Function(
+    Pointer<Float>, Int32);
+
+typedef SherpaOnnxOfflineTtsGenerateWithCallbackNative
+    = Pointer<SherpaOnnxGeneratedAudio> Function(
+        Pointer<SherpaOnnxOfflineTts>,
+        Pointer<Utf8>,
+        Int32,
+        Float,
+        Pointer<NativeFunction<SherpaOnnxGeneratedAudioCallbackNative>>);
+
+typedef SherpaOnnxOfflineTtsGenerateWithCallback
+    = Pointer<SherpaOnnxGeneratedAudio> Function(
+        Pointer<SherpaOnnxOfflineTts>,
+        Pointer<Utf8>,
+        int,
+        double,
+        Pointer<NativeFunction<SherpaOnnxGeneratedAudioCallbackNative>>);
+
 typedef CreateOfflineRecognizerNative = Pointer<SherpaOnnxOfflineRecognizer>
     Function(Pointer<SherpaOnnxOfflineRecognizerConfig>);
 
@@ -608,6 +713,16 @@ typedef SherpaOnnxFreeWaveNative = Void Function(Pointer<SherpaOnnxWave>);
 typedef SherpaOnnxFreeWave = void Function(Pointer<SherpaOnnxWave>);
 
 class SherpaOnnxBindings {
+  static SherpaOnnxCreateOfflineTts? createOfflineTts;
+  static SherpaOnnxDestroyOfflineTts? destroyOfflineTts;
+  static SherpaOnnxOfflineTtsSampleRate? offlineTtsSampleRate;
+  static SherpaOnnxOfflineTtsNumSpeakers? offlineTtsNumSpeakers;
+  static SherpaOnnxOfflineTtsGenerate? offlineTtsGenerate;
+  static SherpaOnnxDestroyOfflineTtsGeneratedAudio?
+      destroyOfflineTtsGeneratedAudio;
+  static SherpaOnnxOfflineTtsGenerateWithCallback?
+      offlineTtsGenerateWithCallback;
+
   static CreateOfflineRecognizer? createOfflineRecognizer;
   static DestroyOfflineRecognizer? destroyOfflineRecognizer;
   static CreateOfflineStream? createOfflineStream;
@@ -740,6 +855,43 @@ class SherpaOnnxBindings {
   static SherpaOnnxFreeWave? freeWave;
 
   static void init(DynamicLibrary dynamicLibrary) {
+    createOfflineTts ??= dynamicLibrary
+        .lookup<NativeFunction<SherpaOnnxCreateOfflineTtsNative>>(
+            'SherpaOnnxCreateOfflineTts')
+        .asFunction();
+
+    destroyOfflineTts ??= dynamicLibrary
+        .lookup<NativeFunction<SherpaOnnxDestroyOfflineTtsNative>>(
+            'SherpaOnnxDestroyOfflineTts')
+        .asFunction();
+
+    offlineTtsSampleRate ??= dynamicLibrary
+        .lookup<NativeFunction<SherpaOnnxOfflineTtsSampleRateNative>>(
+            'SherpaOnnxOfflineTtsSampleRate')
+        .asFunction();
+
+    offlineTtsNumSpeakers ??= dynamicLibrary
+        .lookup<NativeFunction<SherpaOnnxOfflineTtsNumSpeakersNative>>(
+            'SherpaOnnxOfflineTtsNumSpeakers')
+        .asFunction();
+
+    offlineTtsGenerate ??= dynamicLibrary
+        .lookup<NativeFunction<SherpaOnnxOfflineTtsGenerateNative>>(
+            'SherpaOnnxOfflineTtsGenerate')
+        .asFunction();
+
+    destroyOfflineTtsGeneratedAudio ??= dynamicLibrary
+        .lookup<
+                NativeFunction<
+                    SherpaOnnxDestroyOfflineTtsGeneratedAudioNative>>(
+            'SherpaOnnxDestroyOfflineTtsGeneratedAudio')
+        .asFunction();
+
+    offlineTtsGenerateWithCallback ??= dynamicLibrary
+        .lookup<NativeFunction<SherpaOnnxOfflineTtsGenerateWithCallbackNative>>(
+            'SherpaOnnxOfflineTtsGenerateWithCallback')
+        .asFunction();
+
     createOfflineRecognizer ??= dynamicLibrary
         .lookup<NativeFunction<CreateOfflineRecognizerNative>>(
             'CreateOfflineRecognizer')

sherpa-onnx/flutter/lib/src/tts.dart

@@ -0,0 +1,195 @@
+// Copyright (c)  2024  Xiaomi Corporation
+import 'dart:convert';
+import 'dart:ffi';
+import 'dart:typed_data';
+
+import 'package:ffi/ffi.dart';
+
+import './sherpa_onnx_bindings.dart';
+
+class OfflineTtsVitsModelConfig {
+  const OfflineTtsVitsModelConfig({
+    required this.model,
+    this.lexicon = '',
+    required this.tokens,
+    this.dataDir = '',
+    this.noiseScale = 0.667,
+    this.noiseScaleW = 0.8,
+    this.lengthScale = 1.0,
+    this.dictDir = '',
+  });
+
+  @override
+  String toString() {
+    return 'OfflineTtsVitsModelConfig(model: $model, lexicon: $lexicon, tokens: $tokens, dataDir: $dataDir, noiseScale: $noiseScale, noiseScaleW: $noiseScaleW, lengthScale: $lengthScale, dictDir: $dictDir)';
+  }
+
+  final String model;
+  final String lexicon;
+  final String tokens;
+  final String dataDir;
+  final double noiseScale;
+  final double noiseScaleW;
+  final double lengthScale;
+  final String dictDir;
+}
+
+class OfflineTtsModelConfig {
+  const OfflineTtsModelConfig({
+    required this.vits,
+    this.numThreads = 1,
+    this.debug = true,
+    this.provider = 'cpu',
+  });
+
+  @override
+  String toString() {
+    return 'OfflineTtsModelConfig(vits: $vits, numThreads: $numThreads, debug: $debug, provider: $provider)';
+  }
+
+  final OfflineTtsVitsModelConfig vits;
+  final int numThreads;
+  final bool debug;
+  final String provider;
+}
+
+class OfflineTtsConfig {
+  const OfflineTtsConfig({
+    required this.model,
+    this.ruleFsts = '',
+    this.maxNumSenetences = 1,
+    this.ruleFars = '',
+  });
+
+  @override
+  String toString() {
+    return 'OfflineTtsConfig(model: $model, ruleFsts: $ruleFsts, maxNumSenetences: $maxNumSenetences, ruleFars: $ruleFars)';
+  }
+
+  final OfflineTtsModelConfig model;
+  final String ruleFsts;
+  final int maxNumSenetences;
+  final String ruleFars;
+}
+
+class GeneratedAudio {
+  GeneratedAudio({
+    required this.samples,
+    required this.sampleRate,
+  });
+
+  final Float32List samples;
+  final int sampleRate;
+}
+
+class OfflineTts {
+  OfflineTts._({required this.ptr, required this.config});
+
+  /// The user is responsible to call the OfflineTts.free()
+  /// method of the returned instance to avoid memory leak.
+  factory OfflineTts(OfflineTtsConfig config) {
+    final c = calloc<SherpaOnnxOfflineTtsConfig>();
+    c.ref.model.vits.model = config.model.vits.model.toNativeUtf8();
+    c.ref.model.vits.lexicon = config.model.vits.lexicon.toNativeUtf8();
+    c.ref.model.vits.tokens = config.model.vits.tokens.toNativeUtf8();
+    c.ref.model.vits.dataDir = config.model.vits.dataDir.toNativeUtf8();
+    c.ref.model.vits.noiseScale = config.model.vits.noiseScale;
+    c.ref.model.vits.noiseScaleW = config.model.vits.noiseScaleW;
+    c.ref.model.vits.lengthScale = config.model.vits.lengthScale;
+    c.ref.model.vits.dictDir = config.model.vits.dictDir.toNativeUtf8();
+
+    c.ref.model.numThreads = config.model.numThreads;
+    c.ref.model.debug = config.model.debug ? 1 : 0;
+    c.ref.model.provider = config.model.provider.toNativeUtf8();
+
+    c.ref.ruleFsts = config.ruleFsts.toNativeUtf8();
+    c.ref.maxNumSenetences = config.maxNumSenetences;
+    c.ref.ruleFars = config.ruleFars.toNativeUtf8();
+
+    final ptr = SherpaOnnxBindings.createOfflineTts?.call(c) ?? nullptr;
+
+    calloc.free(c.ref.ruleFars);
+    calloc.free(c.ref.ruleFsts);
+    calloc.free(c.ref.model.provider);
+    calloc.free(c.ref.model.vits.dictDir);
+    calloc.free(c.ref.model.vits.dataDir);
+    calloc.free(c.ref.model.vits.tokens);
+    calloc.free(c.ref.model.vits.lexicon);
+    calloc.free(c.ref.model.vits.model);
+
+    return OfflineTts._(ptr: ptr, config: config);
+  }
+
+  void free() {
+    SherpaOnnxBindings.destroyOfflineTts?.call(ptr);
+    ptr = nullptr;
+  }
+
+  GeneratedAudio generate(
+      {required String text, int sid = 0, double speed = 1.0}) {
+    final Pointer<Utf8> textPtr = text.toNativeUtf8();
+    final p =
+        SherpaOnnxBindings.offlineTtsGenerate?.call(ptr, textPtr, sid, speed) ??
+            nullptr;
+    calloc.free(textPtr);
+
+    if (p == nullptr) {
+      return GeneratedAudio(samples: Float32List(0), sampleRate: 0);
+    }
+
+    final samples = p.ref.samples.asTypedList(p.ref.n);
+    final sampleRate = p.ref.sampleRate;
+    final newSamples = Float32List.fromList(samples);
+
+    SherpaOnnxBindings.destroyOfflineTtsGeneratedAudio?.call(p);
+
+    return GeneratedAudio(samples: newSamples, sampleRate: sampleRate);
+  }
+
+  GeneratedAudio generateWithCallback(
+      {required String text,
+      int sid = 0,
+      double speed = 1.0,
+      required void Function(Float32List samples) callback}) {
+    // see
+    // https://github.com/dart-lang/sdk/issues/54276#issuecomment-1846109285
+    // https://stackoverflow.com/questions/69537440/callbacks-in-dart-dartffi-only-supports-calling-static-dart-functions-from-nat
+    // https://github.com/dart-lang/sdk/blob/main/tests/ffi/isolate_local_function_callbacks_test.dart#L46
+    final wrapper =
+        NativeCallable<SherpaOnnxGeneratedAudioCallbackNative>.isolateLocal(
+            (Pointer<Float> samples, int n) {
+      final s = samples.asTypedList(n);
+      final newSamples = Float32List.fromList(s);
+      callback(newSamples);
+    });
+
+    final Pointer<Utf8> textPtr = text.toNativeUtf8();
+    final p = SherpaOnnxBindings.offlineTtsGenerateWithCallback
+            ?.call(ptr, textPtr, sid, speed, wrapper.nativeFunction) ??
+        nullptr;
+
+    calloc.free(textPtr);
+    wrapper.close();
+
+    if (p == nullptr) {
+      return GeneratedAudio(samples: Float32List(0), sampleRate: 0);
+    }
+
+    final samples = p.ref.samples.asTypedList(p.ref.n);
+    final sampleRate = p.ref.sampleRate;
+    final newSamples = Float32List.fromList(samples);
+
+    SherpaOnnxBindings.destroyOfflineTtsGeneratedAudio?.call(p);
+
+    return GeneratedAudio(samples: newSamples, sampleRate: sampleRate);
+  }
+
+  int get sampleRate =>
+      SherpaOnnxBindings.offlineTtsSampleRate?.call(this.ptr) ?? 0;
+
+  int get numSpeakers =>
+      SherpaOnnxBindings.offlineTtsNumSpeakers?.call(this.ptr) ?? 0;
+
+  Pointer<SherpaOnnxOfflineTts> ptr;
+  OfflineTtsConfig config;
+}