Add Dart API for KittenTTS (#2475)

@@ -4,6 +4,34 @@ set -ex
 
 cd dart-api-examples
 
+pushd tts
+
+echo '----------matcha tts----------'
+./run-kitten-en.sh
+./run-kokoro-zh-en.sh
+./run-kokoro-en.sh
+./run-matcha-zh.sh
+./run-matcha-en.sh
+ls -lh *.wav
+rm -rf matcha-icefall-*
+rm *.onnx
+
+echo '----------piper tts----------'
+./run-piper.sh
+rm -rf vits-piper-*
+
+echo '----------coqui tts----------'
+./run-coqui.sh
+rm -rf vits-coqui-*
+
+echo '----------zh tts----------'
+./run-vits-zh.sh
+rm -rf sherpa-onnx-*
+
+ls -lh *.wav
+
+popd # tts
+
 pushd vad
 ./run-ten-vad.sh
 ./run.sh
@@ -72,33 +100,6 @@ echo "speech enhancement with gtcrn models"
 ls -lh
 popd
 
-pushd tts
-
-echo '----------matcha tts----------'
-./run-kokoro-zh-en.sh
-./run-kokoro-en.sh
-./run-matcha-zh.sh
-./run-matcha-en.sh
-ls -lh *.wav
-rm -rf matcha-icefall-*
-rm *.onnx
-
-echo '----------piper tts----------'
-./run-piper.sh
-rm -rf vits-piper-*
-
-echo '----------coqui tts----------'
-./run-coqui.sh
-rm -rf vits-coqui-*
-
-echo '----------zh tts----------'
-./run-vits-zh.sh
-rm -rf sherpa-onnx-*
-
-ls -lh *.wav
-
-popd # tts
-
 pushd speaker-diarization
 echo '----------speaker diarization----------'
 ./run.sh

+// Copyright (c)  2025  Xiaomi Corporation
+import 'dart:io';
+
+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('voices', help: 'Path to the voices.bin')
+    ..addOption('tokens', help: 'Path to tokens.txt')
+    ..addOption(
+      'data-dir',
+      help: 'Path to espeak-ng-data 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['voices'] == 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 voices = res['voices'] as String;
+  final tokens = res['tokens'] as String;
+  final dataDir = res['data-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 kitten = sherpa_onnx.OfflineTtsKittenModelConfig(
+    model: model,
+    voices: voices,
+    tokens: tokens,
+    dataDir: dataDir,
+    lengthScale: 1 / speed,
+  );
+
+  final modelConfig = sherpa_onnx.OfflineTtsModelConfig(
+    kitten: kitten,
+    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');
+}

+#!/usr/bin/env bash
+
+set -ex
+
+dart pub get
+
+# please visit
+# https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/kitten.html
+# to download more models
+if [ ! -f ./kitten-nano-en-v0_1-fp16/model.fp16.onnx ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kitten-nano-en-v0_1-fp16.tar.bz2
+  tar xf kitten-nano-en-v0_1-fp16.tar.bz2
+  rm kitten-nano-en-v0_1-fp16.tar.bz2
+fi
+
+dart run \
+  ./bin/kitten-en.dart \
+  --model ./kitten-nano-en-v0_1-fp16/model.fp16.onnx \
+  --voices ./kitten-nano-en-v0_1-fp16/voices.bin \
+  --tokens ./kitten-nano-en-v0_1-fp16/tokens.txt \
+  --data-dir ./kitten-nano-en-v0_1-fp16/espeak-ng-data \
+  --sid 0 \
+  --speed 1.0 \
+  --output-wav kitten-en-0.wav \
+  --text "Friends fell out often because life was changing so fast. The easiest thing in the world was to lose touch with someone."
+
+ls -lh *.wav

@@ -150,9 +150,18 @@ class AudioTagging {
     final labelsPtr = config.labels.toNativeUtf8();
     c.ref.labels = labelsPtr;
 
+    if (SherpaOnnxBindings.sherpaOnnxCreateAudioTagging == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr =
         SherpaOnnxBindings.sherpaOnnxCreateAudioTagging?.call(c) ?? nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception(
+          "Failed to create audio tagging. Please check your config");
+    }
+
     calloc.free(labelsPtr);
     calloc.free(providerPtr);
     calloc.free(cedPtr);

@@ -140,8 +140,16 @@ class KeywordSpotter {
     c.ref.keywordsBuf = config.keywordsBuf.toNativeUtf8();
     c.ref.keywordsBufSize = config.keywordsBufSize;
 
+    if (SherpaOnnxBindings.createKeywordSpotter == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr = SherpaOnnxBindings.createKeywordSpotter?.call(c) ?? nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception("Failed to create kws. Please check your config");
+    }
+
     calloc.free(c.ref.keywordsBuf);
     calloc.free(c.ref.keywordsFile);
     calloc.free(c.ref.model.bpeVocab);

@@ -79,10 +79,19 @@ class OfflinePunctuation {
     final providerPtr = config.model.provider.toNativeUtf8();
     c.ref.model.provider = providerPtr;
 
+    if (SherpaOnnxBindings.sherpaOnnxCreateOfflinePunctuation == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr =
         SherpaOnnxBindings.sherpaOnnxCreateOfflinePunctuation?.call(c) ??
             nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception(
+          "Failed to create offline punctuation. Please check your config");
+    }
+
     calloc.free(providerPtr);
     calloc.free(ctTransformerPtr);
     calloc.free(c);

@@ -598,8 +598,17 @@ class OfflineRecognizer {
   factory OfflineRecognizer(OfflineRecognizerConfig config) {
     final c = convertConfig(config);
 
+    if (SherpaOnnxBindings.createOfflineRecognizer == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr = SherpaOnnxBindings.createOfflineRecognizer?.call(c) ?? nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception(
+          "Failed to create offline recognizer. Please check your config");
+    }
+
     freeConfig(c);
 
     return OfflineRecognizer._(ptr: ptr, config: config);

@@ -211,10 +211,19 @@ class OfflineSpeakerDiarization {
     c.ref.minDurationOn = config.minDurationOn;
     c.ref.minDurationOff = config.minDurationOff;
 
+    if (SherpaOnnxBindings.sherpaOnnxCreateOfflineSpeakerDiarization == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr =
         SherpaOnnxBindings.sherpaOnnxCreateOfflineSpeakerDiarization?.call(c) ??
             nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception(
+          "Failed to create offline speaker diarization. Please check your config");
+    }
+
     calloc.free(c.ref.embedding.provider);
     calloc.free(c.ref.embedding.model);
     calloc.free(c.ref.segmentation.provider);

@@ -118,10 +118,19 @@ class OfflineSpeechDenoiser {
     c.ref.model.debug = config.model.debug ? 1 : 0;
     c.ref.model.provider = config.model.provider.toNativeUtf8();
 
+    if (SherpaOnnxBindings.sherpaOnnxCreateOfflineSpeechDenoiser == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr =
         SherpaOnnxBindings.sherpaOnnxCreateOfflineSpeechDenoiser?.call(c) ??
             nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception(
+          "Failed to create offline speech denoiser. Please check your config");
+    }
+
     calloc.free(c.ref.model.provider);
     calloc.free(c.ref.model.gtcrn.model);
 

@@ -89,9 +89,18 @@ class OnlinePunctuation {
     final providerPtr = config.model.provider.toNativeUtf8();
     c.ref.model.provider = providerPtr;
 
+    if (SherpaOnnxBindings.sherpaOnnxCreateOnlinePunctuation == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr = SherpaOnnxBindings.sherpaOnnxCreateOnlinePunctuation?.call(c) ??
         nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception(
+          "Failed to create online punctuation. Please check your config");
+    }
+
     // Free the allocated strings and struct memory
     calloc.free(providerPtr);
     calloc.free(cnnBiLstmPtr);

@@ -391,8 +391,17 @@ class OnlineRecognizer {
     c.ref.hr.lexicon = config.hr.lexicon.toNativeUtf8();
     c.ref.hr.ruleFsts = config.hr.ruleFsts.toNativeUtf8();
 
+    if (SherpaOnnxBindings.createOnlineRecognizer == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr = SherpaOnnxBindings.createOnlineRecognizer?.call(c) ?? nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception(
+          "Failed to create online recognizer. Please check your config");
+    }
+
     calloc.free(c.ref.hr.dictDir);
     calloc.free(c.ref.hr.lexicon);
     calloc.free(c.ref.hr.ruleFsts);

@@ -204,6 +204,16 @@ final class SherpaOnnxOfflineTtsKokoroModelConfig extends Struct {
   external Pointer<Utf8> lang;
 }
 
+final class SherpaOnnxOfflineTtsKittenModelConfig extends Struct {
+  external Pointer<Utf8> model;
+  external Pointer<Utf8> voices;
+  external Pointer<Utf8> tokens;
+  external Pointer<Utf8> dataDir;
+
+  @Float()
+  external double lengthScale;
+}
+
 final class SherpaOnnxOfflineTtsModelConfig extends Struct {
   external SherpaOnnxOfflineTtsVitsModelConfig vits;
   @Int32()
@@ -215,6 +225,7 @@ final class SherpaOnnxOfflineTtsModelConfig extends Struct {
   external Pointer<Utf8> provider;
   external SherpaOnnxOfflineTtsMatchaModelConfig matcha;
   external SherpaOnnxOfflineTtsKokoroModelConfig kokoro;
+  external SherpaOnnxOfflineTtsKittenModelConfig kitten;
 }
 
 final class SherpaOnnxOfflineTtsConfig extends Struct {

@@ -60,9 +60,18 @@ class SpeakerEmbeddingExtractor {
     final providerPtr = config.provider.toNativeUtf8();
     c.ref.provider = providerPtr;
 
+    if (SherpaOnnxBindings.createSpeakerEmbeddingExtractor == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr =
         SherpaOnnxBindings.createSpeakerEmbeddingExtractor?.call(c) ?? nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception(
+          "Failed to create speaker embedding extractor. Please check your config");
+    }
+
     calloc.free(providerPtr);
     calloc.free(modelPtr);
     calloc.free(c);

@@ -159,11 +159,51 @@ class OfflineTtsKokoroModelConfig {
   final String lang;
 }
 
+class OfflineTtsKittenModelConfig {
+  const OfflineTtsKittenModelConfig({
+    this.model = '',
+    this.voices = '',
+    this.tokens = '',
+    this.dataDir = '',
+    this.lengthScale = 1.0,
+  });
+
+  factory OfflineTtsKittenModelConfig.fromJson(Map<String, dynamic> json) {
+    return OfflineTtsKittenModelConfig(
+      model: json['model'] as String? ?? '',
+      voices: json['voices'] as String? ?? '',
+      tokens: json['tokens'] as String? ?? '',
+      dataDir: json['dataDir'] as String? ?? '',
+      lengthScale: (json['lengthScale'] as num?)?.toDouble() ?? 1.0,
+    );
+  }
+
+  @override
+  String toString() {
+    return 'OfflineTtsKittenModelConfig(model: $model, voices: $voices, tokens: $tokens, dataDir: $dataDir, lengthScale: $lengthScale)';
+  }
+
+  Map<String, dynamic> toJson() => {
+        'model': model,
+        'voices': voices,
+        'tokens': tokens,
+        'dataDir': dataDir,
+        'lengthScale': lengthScale,
+      };
+
+  final String model;
+  final String voices;
+  final String tokens;
+  final String dataDir;
+  final double lengthScale;
+}
+
 class OfflineTtsModelConfig {
   const OfflineTtsModelConfig({
     this.vits = const OfflineTtsVitsModelConfig(),
     this.matcha = const OfflineTtsMatchaModelConfig(),
     this.kokoro = const OfflineTtsKokoroModelConfig(),
+    this.kitten = const OfflineTtsKittenModelConfig(),
     this.numThreads = 1,
     this.debug = true,
     this.provider = 'cpu',
@@ -177,6 +217,8 @@ class OfflineTtsModelConfig {
           json['matcha'] as Map<String, dynamic>? ?? const {}),
       kokoro: OfflineTtsKokoroModelConfig.fromJson(
           json['kokoro'] as Map<String, dynamic>? ?? const {}),
+      kitten: OfflineTtsKittenModelConfig.fromJson(
+          json['kitten'] as Map<String, dynamic>? ?? const {}),
       numThreads: json['numThreads'] as int? ?? 1,
       debug: json['debug'] as bool? ?? true,
       provider: json['provider'] as String? ?? 'cpu',
@@ -185,13 +227,14 @@ class OfflineTtsModelConfig {
 
   @override
   String toString() {
-    return 'OfflineTtsModelConfig(vits: $vits, matcha: $matcha, kokoro: $kokoro, numThreads: $numThreads, debug: $debug, provider: $provider)';
+    return 'OfflineTtsModelConfig(vits: $vits, matcha: $matcha, kokoro: $kokoro, kitten: $kitten, numThreads: $numThreads, debug: $debug, provider: $provider)';
   }
 
   Map<String, dynamic> toJson() => {
         'vits': vits.toJson(),
         'matcha': matcha.toJson(),
         'kokoro': kokoro.toJson(),
+        'kitten': kitten.toJson(),
         'numThreads': numThreads,
         'debug': debug,
         'provider': provider,
@@ -200,6 +243,7 @@ class OfflineTtsModelConfig {
   final OfflineTtsVitsModelConfig vits;
   final OfflineTtsMatchaModelConfig matcha;
   final OfflineTtsKokoroModelConfig kokoro;
+  final OfflineTtsKittenModelConfig kitten;
   final int numThreads;
   final bool debug;
   final String provider;
@@ -292,6 +336,12 @@ class OfflineTts {
     c.ref.model.kokoro.lexicon = config.model.kokoro.lexicon.toNativeUtf8();
     c.ref.model.kokoro.lang = config.model.kokoro.lang.toNativeUtf8();
 
+    c.ref.model.kitten.model = config.model.kitten.model.toNativeUtf8();
+    c.ref.model.kitten.voices = config.model.kitten.voices.toNativeUtf8();
+    c.ref.model.kitten.tokens = config.model.kitten.tokens.toNativeUtf8();
+    c.ref.model.kitten.dataDir = config.model.kitten.dataDir.toNativeUtf8();
+    c.ref.model.kitten.lengthScale = config.model.kitten.lengthScale;
+
     c.ref.model.numThreads = config.model.numThreads;
     c.ref.model.debug = config.model.debug ? 1 : 0;
     c.ref.model.provider = config.model.provider.toNativeUtf8();
@@ -301,12 +351,25 @@ class OfflineTts {
     c.ref.ruleFars = config.ruleFars.toNativeUtf8();
     c.ref.silenceScale = config.silenceScale;
 
+    if (SherpaOnnxBindings.createOfflineTts == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr = SherpaOnnxBindings.createOfflineTts?.call(c) ?? nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception("Failed to create offline tts. Please check your config");
+    }
+
     calloc.free(c.ref.ruleFars);
     calloc.free(c.ref.ruleFsts);
     calloc.free(c.ref.model.provider);
 
+    calloc.free(c.ref.model.kitten.dataDir);
+    calloc.free(c.ref.model.kitten.tokens);
+    calloc.free(c.ref.model.kitten.voices);
+    calloc.free(c.ref.model.kitten.model);
+
     calloc.free(c.ref.model.kokoro.lang);
     calloc.free(c.ref.model.kokoro.lexicon);
     calloc.free(c.ref.model.kokoro.dictDir);

@@ -153,9 +153,19 @@ class CircularBuffer {
   /// to avoid memory leak.
   factory CircularBuffer({required int capacity}) {
     assert(capacity > 0, 'capacity is $capacity');
+
+    if (SherpaOnnxBindings.createCircularBuffer == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final p =
         SherpaOnnxBindings.createCircularBuffer?.call(capacity) ?? nullptr;
 
+    if (p == nullptr) {
+      throw Exception(
+          "Failed to create circular buffer. Please check your config");
+    }
+
     return CircularBuffer._(ptr: p);
   }
 
@@ -243,10 +253,18 @@ class VoiceActivityDetector {
 
     c.ref.debug = config.debug ? 1 : 0;
 
+    if (SherpaOnnxBindings.createVoiceActivityDetector == null) {
+      throw Exception("Please initialize sherpa-onnx first");
+    }
+
     final ptr = SherpaOnnxBindings.createVoiceActivityDetector
             ?.call(c, bufferSizeInSeconds) ??
         nullptr;
 
+    if (ptr == nullptr) {
+      throw Exception("Failed to create vad. Please check your config");
+    }
+
     calloc.free(providerPtr);
     calloc.free(tenVadModelPtr);
     calloc.free(sileroVadModelPtr);

@@ -15,6 +15,11 @@ class WaveData {
 
 WaveData readWave(String filename) {
   final Pointer<Utf8> str = filename.toNativeUtf8();
+
+  if (SherpaOnnxBindings.readWave == null) {
+    throw Exception("Please initialize sherpa-onnx first");
+  }
+
   Pointer<SherpaOnnxWave> wave =
       SherpaOnnxBindings.readWave?.call(str) ?? nullptr;
   calloc.free(str);

@@ -17,6 +17,10 @@ bool writeWave(
   final pList = p.asTypedList(n);
   pList.setAll(0, samples);
 
+  if (SherpaOnnxBindings.writeWave == null) {
+    throw Exception("Please initialize sherpa-onnx first");
+  }
+
   int ok =
       SherpaOnnxBindings.writeWave?.call(p, n, sampleRate, filenamePtr) ?? 0;