| 1 | import 'dart:async'; |
| 2 | import 'dart:io'; |
| 3 | import 'dart:isolate'; |
| 4 | import 'dart:typed_data'; |
| 5 | |
| 6 | import 'package:audio_session/audio_session.dart'; |
| 7 | import 'package:file_picker/file_picker.dart'; |
| 8 | import 'package:just_audio/just_audio.dart'; |
| 9 | import 'package:just_audio_background/just_audio_background.dart'; |
| 10 | import 'package:flutter_riverpod/flutter_riverpod.dart'; |
| 11 | import 'package:flutter_tts/flutter_tts.dart'; |
| 12 | |
| 13 | import '../../../core/logger/app_logger.dart'; |
| 14 | import '../../../core/models/enums.dart'; |
| 15 | import '../../../core/providers/app_providers.dart'; |
| 16 | import '../../../core/utils/safe_file_picker.dart'; |
| 17 | import '../utils/tts_text_processor.dart'; |
| 18 | import 'tts_model_providers.dart'; |
| 19 | import '../data/piper_tts_model.dart'; |
| 20 | import '../data/system_tts_configuration.dart'; |
| 21 | import 'tts_worker.dart'; |
| 22 | import '../utils/wav_merge.dart'; |
| 23 | |
| 24 | final ttsProvider = NotifierProvider<TtsNotifier, TtsState>(() { |
| 25 | return TtsNotifier(); |
| 26 | }); |
| 27 | |
| 28 | class TtsState { |
| 29 | final bool isSpeaking; |
| 30 | final bool isPaused; |
| 31 | final bool isInitializing; |
| 32 | final String? error; |
| 33 | final EngineId? activeEngine; |
| 34 | final String? playingContent; |
| 35 | final String? playingMessageId; |
| 36 | final String? playingConversationId; |
| 37 | final bool isPreview; |
| 38 | final Duration position; |
| 39 | final Duration duration; |
| 40 | final bool canSeek; |
| 41 | final bool isSeekPending; |
| 42 | final double playbackSpeed; |
| 43 | |
| 44 | /// Text chunks currently being read aloud, in order. Cleared when |
| 45 | /// playback stops. Useful for word-level progress UI. |
| 46 | final List<String> chunks; |
| 47 | |
| 48 | /// Index of the chunk currently being read. -1 when nothing is playing. |
| 49 | final int currentChunkIndex; |
| 50 | |
| 51 | /// Character offset currently reported by a TTS engine with native |
| 52 | /// word-progress support. Falls back to an estimated offset for engines |
| 53 | /// that only expose playback time. |
| 54 | final int spokenCharOffset; |
| 55 | |
| 56 | const TtsState({ |
| 57 | this.isSpeaking = false, |
| 58 | this.isPaused = false, |
| 59 | this.isInitializing = false, |
| 60 | this.error, |
| 61 | this.activeEngine, |
| 62 | this.playingContent, |
| 63 | this.playingMessageId, |
| 64 | this.playingConversationId, |
| 65 | this.isPreview = false, |
| 66 | this.position = Duration.zero, |
| 67 | this.duration = Duration.zero, |
| 68 | this.canSeek = false, |
| 69 | this.isSeekPending = false, |
| 70 | this.playbackSpeed = 1.0, |
| 71 | this.chunks = const [], |
| 72 | this.currentChunkIndex = -1, |
| 73 | this.spokenCharOffset = -1, |
| 74 | }); |
| 75 | |
| 76 | TtsState copyWith({ |
| 77 | bool? isSpeaking, |
| 78 | bool? isPaused, |
| 79 | bool? isInitializing, |
| 80 | String? error, |
| 81 | EngineId? activeEngine, |
| 82 | String? playingContent, |
| 83 | String? playingMessageId, |
| 84 | String? playingConversationId, |
| 85 | bool? isPreview, |
| 86 | Duration? position, |
| 87 | Duration? duration, |
| 88 | bool? canSeek, |
| 89 | bool? isSeekPending, |
| 90 | double? playbackSpeed, |
| 91 | bool clearPlayingTarget = false, |
| 92 | bool resetProgress = false, |
| 93 | List<String>? chunks, |
| 94 | int? currentChunkIndex, |
| 95 | int? spokenCharOffset, |
| 96 | }) { |
| 97 | return TtsState( |
| 98 | isSpeaking: isSpeaking ?? this.isSpeaking, |
| 99 | isPaused: isPaused ?? this.isPaused, |
| 100 | isInitializing: isInitializing ?? this.isInitializing, |
| 101 | error: error, |
| 102 | activeEngine: activeEngine ?? this.activeEngine, |
| 103 | playingContent: playingContent, |
| 104 | playingMessageId: clearPlayingTarget |
| 105 | ? null |
| 106 | : (playingMessageId ?? this.playingMessageId), |
| 107 | playingConversationId: clearPlayingTarget |
| 108 | ? null |
| 109 | : (playingConversationId ?? this.playingConversationId), |
| 110 | isPreview: isPreview ?? this.isPreview, |
| 111 | position: resetProgress ? Duration.zero : (position ?? this.position), |
| 112 | duration: resetProgress ? Duration.zero : (duration ?? this.duration), |
| 113 | canSeek: canSeek ?? this.canSeek, |
| 114 | isSeekPending: isSeekPending ?? this.isSeekPending, |
| 115 | playbackSpeed: playbackSpeed ?? this.playbackSpeed, |
| 116 | chunks: chunks ?? this.chunks, |
| 117 | currentChunkIndex: currentChunkIndex ?? this.currentChunkIndex, |
| 118 | spokenCharOffset: resetProgress |
| 119 | ? -1 |
| 120 | : (spokenCharOffset ?? this.spokenCharOffset), |
| 121 | ); |
| 122 | } |
| 123 | } |
| 124 | |
| 125 | class TtsNotifier extends Notifier<TtsState> { |
| 126 | FlutterTts? _flutterTts; |
| 127 | final _systemTtsConfiguration = SystemTtsConfiguration(); |
| 128 | final AudioPlayer _player = AudioPlayer(); |
| 129 | bool _isPlayerDisposed = false; |
| 130 | bool _isStopping = false; |
| 131 | List<String> _chunks = []; |
| 132 | int _currentChunkIndex = 0; |
| 133 | EngineId? _currentEngine; |
| 134 | String? _currentVoice; |
| 135 | String? _currentModelPath; |
| 136 | String? _currentVoicesPath; |
| 137 | String? _currentTokensPath; |
| 138 | String? _currentDataDir; |
| 139 | String? _currentLexicon; |
| 140 | |
| 141 | // Isolate worker fields |
| 142 | Isolate? _workerIsolate; |
| 143 | SendPort? _workerSendPort; |
| 144 | final ReceivePort _mainReceivePort = ReceivePort(); |
| 145 | final Map<int, Uint8List> _synthesizedChunks = {}; |
| 146 | int _lastEnqueuedChunk = -1; |
| 147 | static const int _lookAheadCount = 2; |
| 148 | Completer<void>? _initCompleter; |
| 149 | bool _isWorkerInitializing = false; |
| 150 | int _currentSessionId = 0; |
| 151 | |
| 152 | StreamSubscription<void>? _playerCompleteSubscription; |
| 153 | int _playbackSessionId = 0; |
| 154 | Completer<SendPort>? _workerPortCompleter; |
| 155 | |
| 156 | bool _isPreview = false; |
| 157 | final Map<int, AudioSource> _playlistBuffer = {}; |
| 158 | int _nextPlaylistIndexToAdd = 0; |
| 159 | StreamSubscription<int?>? _currentIndexSubscription; |
| 160 | StreamSubscription<PlaybackEvent>? _playerEventSubscription; |
| 161 | StreamSubscription<Duration>? _positionSubscription; |
| 162 | StreamSubscription<Duration?>? _durationSubscription; |
| 163 | Timer? _systemProgressTimer; |
| 164 | |
| 165 | // Progress tracking |
| 166 | final Map<int, Duration> _chunkDurations = {}; |
| 167 | Duration _totalDuration = Duration.zero; |
| 168 | Duration _chunkOffset = Duration.zero; |
| 169 | final StreamController<void> _onChunkChanged = |
| 170 | StreamController<void>.broadcast(); |
| 171 | |
| 172 | // System TTS approximate seek |
| 173 | String? _systemFullText; |
| 174 | Duration _systemSkippedOffset = Duration.zero; |
| 175 | DateTime? _systemSpeakStart; |
| 176 | bool _systemHasNativeProgress = false; |
| 177 | int _systemSpeakGeneration = 0; |
| 178 | |
| 179 | _MessageAudioCache? _messageAudioCache; |
| 180 | |
| 181 | static const playbackSpeedOptions = [0.5, 0.75, 1.0, 1.25, 1.5, 2.0, 3.0]; |
| 182 | |
| 183 | AudioPlayer get player => _player; |
| 184 | Stream<Duration> get onPositionChanged => _player.positionStream; |
| 185 | Duration get totalDuration => _totalDuration; |
| 186 | Duration get chunkOffset => _chunkOffset; |
| 187 | Stream<void> get onChunkChanged => _onChunkChanged.stream; |
| 188 | |
| 189 | @override |
| 190 | TtsState build() { |
| 191 | ref.keepAlive(); |
| 192 | |
| 193 | ref.listen(settingsProvider, (previous, next) { |
| 194 | if (previous?.ttsEngine != next.ttsEngine) { |
| 195 | state = state.copyWith(activeEngine: next.ttsEngine); |
| 196 | } |
| 197 | }); |
| 198 | |
| 199 | _playerCompleteSubscription = _player.playerStateStream.listen(( |
| 200 | playerState, |
| 201 | ) { |
| 202 | if (playerState.processingState != ProcessingState.completed) return; |
| 203 | if (_isStopping || state.isInitializing) return; |
| 204 | if (_currentSessionId != _playbackSessionId) return; |
| 205 | if (_nextPlaylistIndexToAdd < _chunks.length) return; |
| 206 | if (_lastEnqueuedChunk < _chunks.length - 1) return; |
| 207 | _onPlaybackFinished(); |
| 208 | }); |
| 209 | |
| 210 | _playerEventSubscription = _player.playbackEventStream.listen( |
| 211 | (_) {}, |
| 212 | onError: (Object error) { |
| 213 | Log.warning('TTS player event error: $error'); |
| 214 | if (state.isSpeaking && !_isStopping) { |
| 215 | _onPlaybackFinished(); |
| 216 | } |
| 217 | }, |
| 218 | ); |
| 219 | |
| 220 | _currentIndexSubscription = _player.currentIndexStream.listen((index) { |
| 221 | if (index != null && index != _currentChunkIndex) { |
| 222 | _currentChunkIndex = index; |
| 223 | state = state.copyWith(currentChunkIndex: index); |
| 224 | _updateChunkOffset(); |
| 225 | _onChunkChanged.add(null); |
| 226 | _enqueueNextChunks(); |
| 227 | _updatePlaybackProgress(); |
| 228 | } |
| 229 | }); |
| 230 | |
| 231 | _positionSubscription = _player.positionStream.listen((_) { |
| 232 | _updatePlaybackProgress(); |
| 233 | }); |
| 234 | |
| 235 | _durationSubscription = _player.durationStream.listen((_) { |
| 236 | _updatePlaybackProgress(); |
| 237 | }); |
| 238 | |
| 239 | _setupWorkerListener(); |
| 240 | |
| 241 | ref.onDispose(() { |
| 242 | _workerPortCompleter?.completeError(StateError('Disposed')); |
| 243 | _workerPortCompleter = null; |
| 244 | _workerIsolate?.kill(); |
| 245 | _mainReceivePort.close(); |
| 246 | _playerCompleteSubscription?.cancel(); |
| 247 | _playerEventSubscription?.cancel(); |
| 248 | _currentIndexSubscription?.cancel(); |
| 249 | _positionSubscription?.cancel(); |
| 250 | _durationSubscription?.cancel(); |
| 251 | _systemProgressTimer?.cancel(); |
| 252 | _onChunkChanged.close(); |
| 253 | if (!_isPlayerDisposed) { |
| 254 | _isPlayerDisposed = true; |
| 255 | try { |
| 256 | _player.dispose(); |
| 257 | } catch (_) {} |
| 258 | } |
| 259 | }); |
| 260 | |
| 261 | final settings = ref.read(settingsProvider); |
| 262 | return TtsState(activeEngine: settings.ttsEngine); |
| 263 | } |
| 264 | |
| 265 | void _setupWorkerListener() { |
| 266 | _mainReceivePort.listen((message) async { |
| 267 | if (message is SendPort) { |
| 268 | _workerSendPort = message; |
| 269 | _workerPortCompleter?.complete(message); |
| 270 | _workerPortCompleter = null; |
| 271 | } else if (message is TtsWorkerResponse) { |
| 272 | if (message.isInitResponse) { |
| 273 | _initCompleter?.complete(); |
| 274 | _initCompleter = null; |
| 275 | } else if (message.chunkIndex != null) { |
| 276 | if (message.sessionId != _currentSessionId) { |
| 277 | Log.debug('Ignoring stale chunk ${message.chunkIndex}'); |
| 278 | return; |
| 279 | } |
| 280 | if (message.wavBytes != null) { |
| 281 | _synthesizedChunks[message.chunkIndex!] = message.wavBytes!; |
| 282 | _computeChunkDuration(message.chunkIndex!, message.wavBytes!); |
| 283 | await _addChunkToPlaylist(message.chunkIndex!, message.wavBytes!); |
| 284 | _updatePlaybackProgress(); |
| 285 | } else { |
| 286 | Log.error( |
| 287 | 'Worker error for chunk ${message.chunkIndex}: ${message.error}', |
| 288 | ); |
| 289 | _synthesizedChunks[message.chunkIndex!] = Uint8List(0); |
| 290 | await _addChunkToPlaylist(message.chunkIndex!, Uint8List(0)); |
| 291 | } |
| 292 | } else if (message.error != null) { |
| 293 | Log.error('Worker general error: ${message.error}'); |
| 294 | _initCompleter?.completeError(message.error!); |
| 295 | _initCompleter = null; |
| 296 | } |
| 297 | } |
| 298 | }); |
| 299 | } |
| 300 | |
| 301 | Future<void> _ensureWorker() async { |
| 302 | if (_workerIsolate != null) return; |
| 303 | |
| 304 | Log.info('Spawning TTS worker isolate...'); |
| 305 | _workerPortCompleter = Completer<SendPort>(); |
| 306 | _workerIsolate = await Isolate.spawn( |
| 307 | ttsWorkerEntry, |
| 308 | _mainReceivePort.sendPort, |
| 309 | ); |
| 310 | |
| 311 | await _workerPortCompleter!.future.timeout(const Duration(seconds: 10)); |
| 312 | } |
| 313 | |
| 314 | void _computeChunkDuration(int index, Uint8List wavBytes) { |
| 315 | if (wavBytes.length < 44) return; |
| 316 | final data = ByteData.view(wavBytes.buffer, wavBytes.offsetInBytes, 44); |
| 317 | final sampleRate = data.getUint32(24, Endian.little); |
| 318 | final dataSize = data.getUint32(40, Endian.little); |
| 319 | final sampleCount = dataSize ~/ 2; |
| 320 | final durationMs = (sampleCount * 1000) ~/ sampleRate; |
| 321 | _chunkDurations[index] = Duration(milliseconds: durationMs); |
| 322 | _totalDuration = Duration.zero; |
| 323 | for (final d in _chunkDurations.values) { |
| 324 | _totalDuration += d; |
| 325 | } |
| 326 | _updatePlaybackProgress(); |
| 327 | } |
| 328 | |
| 329 | void _updateChunkOffset() { |
| 330 | var offset = Duration.zero; |
| 331 | for (var i = 0; i < _currentChunkIndex; i++) { |
| 332 | offset += _chunkDurations[i] ?? Duration.zero; |
| 333 | } |
| 334 | _chunkOffset = offset; |
| 335 | } |
| 336 | |
| 337 | Duration _averageChunkDuration() { |
| 338 | if (_chunkDurations.isEmpty) return const Duration(seconds: 3); |
| 339 | final totalMs = _chunkDurations.values.fold<int>( |
| 340 | 0, |
| 341 | (sum, d) => sum + d.inMilliseconds, |
| 342 | ); |
| 343 | return Duration(milliseconds: totalMs ~/ _chunkDurations.length); |
| 344 | } |
| 345 | |
| 346 | Duration _estimatedTotalDuration() { |
| 347 | if (_chunks.isEmpty) return Duration.zero; |
| 348 | if (_chunkDurations.length >= _chunks.length) return _totalDuration; |
| 349 | final avg = _averageChunkDuration(); |
| 350 | return Duration(milliseconds: avg.inMilliseconds * _chunks.length); |
| 351 | } |
| 352 | |
| 353 | Duration _computeGlobalPosition() { |
| 354 | if (state.activeEngine == EngineId.system) { |
| 355 | return _systemEstimatedPosition(); |
| 356 | } |
| 357 | _updateChunkOffset(); |
| 358 | return _chunkOffset + _player.position; |
| 359 | } |
| 360 | |
| 361 | void _updatePlaybackProgress() { |
| 362 | if (!state.isSpeaking || state.isPreview) return; |
| 363 | |
| 364 | if (state.activeEngine == EngineId.system) { |
| 365 | return; |
| 366 | } |
| 367 | |
| 368 | final position = _computeGlobalPosition(); |
| 369 | final duration = _estimatedTotalDuration(); |
| 370 | state = state.copyWith( |
| 371 | position: position, |
| 372 | duration: duration, |
| 373 | canSeek: |
| 374 | !state.isInitializing && |
| 375 | duration > Duration.zero && |
| 376 | _nextPlaylistIndexToAdd >= _chunks.length, |
| 377 | ); |
| 378 | } |
| 379 | |
| 380 | Duration _systemEstimatedDuration() { |
| 381 | if (_systemFullText == null) return Duration.zero; |
| 382 | final cps = 12.0 * ref.read(settingsProvider).ttsSpeed; |
| 383 | return Duration( |
| 384 | milliseconds: (_systemFullText!.length / cps * 1000).round(), |
| 385 | ); |
| 386 | } |
| 387 | |
| 388 | Duration _systemEstimatedPosition() { |
| 389 | if (_systemSpeakStart == null) return _systemSkippedOffset; |
| 390 | final pos = |
| 391 | _systemSkippedOffset + DateTime.now().difference(_systemSpeakStart!); |
| 392 | final total = _systemEstimatedDuration(); |
| 393 | return pos > total ? total : pos; |
| 394 | } |
| 395 | |
| 396 | void _stopSystemProgressTimer() { |
| 397 | _systemProgressTimer?.cancel(); |
| 398 | _systemProgressTimer = null; |
| 399 | } |
| 400 | |
| 401 | void _startSystemProgressTimer(int speakGeneration) { |
| 402 | _stopSystemProgressTimer(); |
| 403 | _systemProgressTimer = Timer.periodic(const Duration(milliseconds: 80), ( |
| 404 | timer, |
| 405 | ) { |
| 406 | if (_isStopping || |
| 407 | speakGeneration != _systemSpeakGeneration || |
| 408 | !state.isSpeaking || |
| 409 | state.activeEngine != EngineId.system) { |
| 410 | timer.cancel(); |
| 411 | if (identical(_systemProgressTimer, timer)) { |
| 412 | _systemProgressTimer = null; |
| 413 | } |
| 414 | return; |
| 415 | } |
| 416 | |
| 417 | final duration = _systemEstimatedDuration(); |
| 418 | final position = _systemEstimatedPosition(); |
| 419 | final text = _systemFullText ?? ''; |
| 420 | int? estimatedCharOffset; |
| 421 | if (!_systemHasNativeProgress && text.isNotEmpty) { |
| 422 | final durationMicros = duration.inMicroseconds; |
| 423 | final progress = durationMicros <= 0 |
| 424 | ? 0.0 |
| 425 | : position.inMicroseconds / durationMicros; |
| 426 | estimatedCharOffset = (progress * text.length).floor().clamp( |
| 427 | 0, |
| 428 | text.length - 1, |
| 429 | ); |
| 430 | } |
| 431 | |
| 432 | state = state.copyWith( |
| 433 | position: position, |
| 434 | duration: duration, |
| 435 | spokenCharOffset: estimatedCharOffset, |
| 436 | playingContent: state.playingContent, |
| 437 | ); |
| 438 | }); |
| 439 | } |
| 440 | |
| 441 | void _resetSystemSeekState() { |
| 442 | _systemFullText = null; |
| 443 | _systemSkippedOffset = Duration.zero; |
| 444 | _systemSpeakStart = null; |
| 445 | _systemHasNativeProgress = false; |
| 446 | _stopSystemProgressTimer(); |
| 447 | } |
| 448 | |
| 449 | void _resetProgress() { |
| 450 | _chunkDurations.clear(); |
| 451 | _totalDuration = Duration.zero; |
| 452 | _chunkOffset = Duration.zero; |
| 453 | _resetSystemSeekState(); |
| 454 | } |
| 455 | |
| 456 | void _deleteChunkFile(int sessionId, int chunkIndex) { |
| 457 | if (chunkIndex < 0) return; |
| 458 | Future.microtask(() async { |
| 459 | try { |
| 460 | final tempFile = File( |
| 461 | '${Directory.systemTemp.path}/tts_chunk_${sessionId}_$chunkIndex.wav', |
| 462 | ); |
| 463 | if (await tempFile.exists()) { |
| 464 | await tempFile.delete(); |
| 465 | } |
| 466 | } catch (_) {} |
| 467 | }); |
| 468 | } |
| 469 | |
| 470 | void _cleanupSessionFiles(int sessionId, int chunkCount) { |
| 471 | for (var i = 0; i < chunkCount; i++) { |
| 472 | _deleteChunkFile(sessionId, i); |
| 473 | } |
| 474 | } |
| 475 | |
| 476 | void _onPlaybackFinished() { |
| 477 | if (_isStopping) return; |
| 478 | if (_currentSessionId != _playbackSessionId) return; |
| 479 | _saveMessageAudioCache(); |
| 480 | _cleanupSessionFiles(_currentSessionId, _chunks.length); |
| 481 | _resetProgress(); |
| 482 | state = state.copyWith( |
| 483 | isSpeaking: false, |
| 484 | isPaused: false, |
| 485 | isInitializing: false, |
| 486 | playingContent: null, |
| 487 | isPreview: false, |
| 488 | clearPlayingTarget: true, |
| 489 | resetProgress: true, |
| 490 | chunks: const [], |
| 491 | currentChunkIndex: -1, |
| 492 | canSeek: false, |
| 493 | ); |
| 494 | // Hand the iOS audio session back to record so a follow-up |
| 495 | // STT startListening() can capture audio immediately. |
| 496 | _configureAudioSessionForSpeech(); |
| 497 | } |
| 498 | |
| 499 | void _configureAudioSessionForSpeech() { |
| 500 | try { |
| 501 | AudioSession.instance.then((session) { |
| 502 | session.configure(const AudioSessionConfiguration.speech()); |
| 503 | }); |
| 504 | } catch (e) { |
| 505 | Log.error('Error configuring audio session for speech: $e'); |
| 506 | } |
| 507 | } |
| 508 | |
| 509 | List<String> _splitText(String text) { |
| 510 | if (text.length <= 250) return [text]; |
| 511 | |
| 512 | final regex = RegExp(r'(?<=[.!?])\s+'); |
| 513 | final sentences = text.split(regex); |
| 514 | |
| 515 | final List<String> chunks = []; |
| 516 | String currentChunk = ""; |
| 517 | |
| 518 | for (final sentence in sentences) { |
| 519 | if ((currentChunk.length + sentence.length) > 250 && |
| 520 | currentChunk.isNotEmpty) { |
| 521 | chunks.add(currentChunk.trim()); |
| 522 | currentChunk = ""; |
| 523 | } |
| 524 | |
| 525 | if (sentence.length > 250) { |
| 526 | final subRegex = RegExp(r'(?<=[,;])\s+'); |
| 527 | final parts = sentence.split(subRegex); |
| 528 | for (final part in parts) { |
| 529 | if ((currentChunk.length + part.length) > 250 && |
| 530 | currentChunk.isNotEmpty) { |
| 531 | chunks.add(currentChunk.trim()); |
| 532 | currentChunk = ""; |
| 533 | } |
| 534 | currentChunk += "$part "; |
| 535 | } |
| 536 | } else { |
| 537 | currentChunk += "$sentence "; |
| 538 | } |
| 539 | } |
| 540 | |
| 541 | if (currentChunk.trim().isNotEmpty) { |
| 542 | chunks.add(currentChunk.trim()); |
| 543 | } |
| 544 | |
| 545 | return chunks; |
| 546 | } |
| 547 | |
| 548 | Future<void> _initEngine(EngineId engineId, {String? voiceId}) async { |
| 549 | await _ensureWorker(); |
| 550 | |
| 551 | if (_isWorkerInitializing && _initCompleter != null) { |
| 552 | await _initCompleter!.future.timeout(const Duration(seconds: 30)); |
| 553 | } |
| 554 | |
| 555 | String modelPath = ''; |
| 556 | String voicesPath = ''; |
| 557 | String tokensPath = ''; |
| 558 | String dataDir = ''; |
| 559 | String lexicon = ''; |
| 560 | |
| 561 | switch (engineId) { |
| 562 | case EngineId.kitten: |
| 563 | final variant = ref.read(settingsProvider).kittenTtsModelVariant; |
| 564 | final kittenDl = ref.read(kittenTtsDownloaderProvider); |
| 565 | final mPath = await kittenDl.getModelPath(variant); |
| 566 | final vPath = await kittenDl.getVoicesPath(variant); |
| 567 | final tPath = await kittenDl.getTokensPath(variant); |
| 568 | final dDir = await kittenDl.getDataDir(variant); |
| 569 | if (mPath == null || vPath == null || tPath == null) { |
| 570 | throw StateError( |
| 571 | 'Kitten model files not found for ${variant.name}. Please download them in the Model Manager.', |
| 572 | ); |
| 573 | } |
| 574 | modelPath = mPath; |
| 575 | voicesPath = vPath; |
| 576 | tokensPath = tPath; |
| 577 | dataDir = dDir.path; |
| 578 | break; |
| 579 | case EngineId.piper: |
| 580 | final selectedVoiceId = |
| 581 | voiceId ?? |
| 582 | ref.read(settingsProvider).ttsVoiceId ?? |
| 583 | piperVoices.first.id; |
| 584 | final requestedVariant = PiperTtsModelVariant.values.firstWhere( |
| 585 | (v) => v.id == selectedVoiceId, |
| 586 | orElse: () => PiperTtsModelVariant.enUsLessacMedium, |
| 587 | ); |
| 588 | final downloader = ref.read(piperTtsDownloaderProvider); |
| 589 | final downloadedVariants = await downloader.getDownloadedVariants(); |
| 590 | final variant = downloadedVariants.contains(requestedVariant) |
| 591 | ? requestedVariant |
| 592 | : downloadedVariants.isNotEmpty |
| 593 | ? downloadedVariants.first |
| 594 | : requestedVariant; |
| 595 | final mPath = await downloader.getModelPath(variant); |
| 596 | final tPath = await downloader.getTokensPath(variant); |
| 597 | final dDir = await downloader.getDataDir(variant); |
| 598 | if (mPath == null || tPath == null) { |
| 599 | throw StateError( |
| 600 | 'Piper voice files not found for ${variant.displayName}. Please download them in the Model Manager.', |
| 601 | ); |
| 602 | } |
| 603 | modelPath = mPath; |
| 604 | tokensPath = tPath; |
| 605 | dataDir = dDir.path; |
| 606 | break; |
| 607 | default: |
| 608 | return; |
| 609 | } |
| 610 | |
| 611 | if (_currentEngine == engineId && |
| 612 | !_isWorkerInitializing && |
| 613 | _currentModelPath == modelPath && |
| 614 | _currentVoicesPath == voicesPath && |
| 615 | _currentTokensPath == tokensPath && |
| 616 | _currentDataDir == dataDir && |
| 617 | _currentLexicon == lexicon) { |
| 618 | return; |
| 619 | } |
| 620 | |
| 621 | _isWorkerInitializing = true; |
| 622 | |
| 623 | try { |
| 624 | _initCompleter = Completer<void>(); |
| 625 | _workerSendPort!.send( |
| 626 | TtsInitRequest( |
| 627 | engineId: engineId, |
| 628 | modelPath: modelPath, |
| 629 | voicesPath: voicesPath, |
| 630 | tokensPath: tokensPath, |
| 631 | dataDir: dataDir, |
| 632 | lexicon: lexicon, |
| 633 | ), |
| 634 | ); |
| 635 | |
| 636 | await _initCompleter!.future.timeout(const Duration(seconds: 30)); |
| 637 | _currentEngine = engineId; |
| 638 | _currentModelPath = modelPath; |
| 639 | _currentVoicesPath = voicesPath; |
| 640 | _currentTokensPath = tokensPath; |
| 641 | _currentDataDir = dataDir; |
| 642 | _currentLexicon = lexicon; |
| 643 | } finally { |
| 644 | _isWorkerInitializing = false; |
| 645 | } |
| 646 | } |
| 647 | |
| 648 | Future<void> speak( |
| 649 | String text, { |
| 650 | EngineId? engineId, |
| 651 | String? voiceId, |
| 652 | bool isPreview = false, |
| 653 | String? messageId, |
| 654 | String? conversationId, |
| 655 | }) async { |
| 656 | if (text.trim().isEmpty) return; |
| 657 | |
| 658 | final settings = ref.read(settingsProvider); |
| 659 | if (settings.ttsProcessMarkdown) { |
| 660 | text = TtsTextProcessor.process(text, stripMarkdown: true); |
| 661 | if (text.trim().isEmpty) return; |
| 662 | } |
| 663 | |
| 664 | engineId ??= settings.ttsEngine; |
| 665 | final resolvedEngine = engineId; |
| 666 | |
| 667 | final availableVoices = voicesForEngine(resolvedEngine); |
| 668 | final resolvedVoiceId = |
| 669 | voiceId ?? |
| 670 | settings.ttsVoiceId ?? |
| 671 | (availableVoices.isNotEmpty ? availableVoices.first.id : null); |
| 672 | |
| 673 | Future<bool> tryPlayCached() async { |
| 674 | if (isPreview || |
| 675 | messageId == null || |
| 676 | _messageAudioCache == null || |
| 677 | _messageAudioCache!.messageId != messageId || |
| 678 | _messageAudioCache!.text != text || |
| 679 | _messageAudioCache!.engine != resolvedEngine || |
| 680 | _messageAudioCache!.voice != resolvedVoiceId) { |
| 681 | return false; |
| 682 | } |
| 683 | await _playFromCache( |
| 684 | isPreview: isPreview, |
| 685 | messageId: messageId, |
| 686 | conversationId: conversationId, |
| 687 | engineId: resolvedEngine, |
| 688 | text: text, |
| 689 | ); |
| 690 | return true; |
| 691 | } |
| 692 | |
| 693 | if (await tryPlayCached()) return; |
| 694 | |
| 695 | _isStopping = false; |
| 696 | if (state.isSpeaking) { |
| 697 | await stop(); |
| 698 | await Future.delayed(const Duration(milliseconds: 100)); |
| 699 | _isStopping = false; |
| 700 | } |
| 701 | |
| 702 | if (await tryPlayCached()) return; |
| 703 | |
| 704 | if (resolvedEngine == EngineId.system) { |
| 705 | _flutterTts ??= FlutterTts(); |
| 706 | _resetSystemSeekState(); |
| 707 | final speakGeneration = ++_systemSpeakGeneration; |
| 708 | _systemFullText = text; |
| 709 | _systemSkippedOffset = Duration.zero; |
| 710 | _systemSpeakStart = null; |
| 711 | _systemHasNativeProgress = false; |
| 712 | final estimatedDuration = _systemEstimatedDuration(); |
| 713 | state = state.copyWith( |
| 714 | isSpeaking: true, |
| 715 | isPaused: false, |
| 716 | isInitializing: false, |
| 717 | error: null, |
| 718 | activeEngine: EngineId.system, |
| 719 | playingContent: text, |
| 720 | playingMessageId: isPreview ? null : messageId, |
| 721 | playingConversationId: isPreview ? null : conversationId, |
| 722 | isPreview: isPreview, |
| 723 | canSeek: false, |
| 724 | position: Duration.zero, |
| 725 | duration: estimatedDuration, |
| 726 | chunks: [text], |
| 727 | currentChunkIndex: 0, |
| 728 | spokenCharOffset: 0, |
| 729 | ); |
| 730 | try { |
| 731 | await _systemTtsConfiguration.prepare( |
| 732 | _flutterTts!, |
| 733 | isAndroid: Platform.isAndroid, |
| 734 | ); |
| 735 | if (_isStopping || speakGeneration != _systemSpeakGeneration) return; |
| 736 | await _applySystemSpeechRate(); |
| 737 | _flutterTts!.setStartHandler(() { |
| 738 | if (_isStopping || speakGeneration != _systemSpeakGeneration) return; |
| 739 | _systemSpeakStart = DateTime.now(); |
| 740 | _startSystemProgressTimer(speakGeneration); |
| 741 | }); |
| 742 | _flutterTts!.setProgressHandler((spokenText, start, end, word) { |
| 743 | if (_isStopping || speakGeneration != _systemSpeakGeneration) return; |
| 744 | if (text.isEmpty) return; |
| 745 | |
| 746 | _systemHasNativeProgress = true; |
| 747 | final safeStart = start.clamp(0, text.length - 1); |
| 748 | final duration = _systemEstimatedDuration(); |
| 749 | final position = Duration( |
| 750 | microseconds: (duration.inMicroseconds * (safeStart / text.length)) |
| 751 | .round(), |
| 752 | ); |
| 753 | state = state.copyWith( |
| 754 | position: position, |
| 755 | duration: duration, |
| 756 | spokenCharOffset: safeStart, |
| 757 | playingContent: state.playingContent, |
| 758 | ); |
| 759 | }); |
| 760 | _flutterTts!.setCompletionHandler(() { |
| 761 | if (_isStopping || speakGeneration != _systemSpeakGeneration) return; |
| 762 | _resetSystemSeekState(); |
| 763 | state = state.copyWith( |
| 764 | isSpeaking: false, |
| 765 | isPaused: false, |
| 766 | playingContent: null, |
| 767 | clearPlayingTarget: true, |
| 768 | resetProgress: true, |
| 769 | canSeek: false, |
| 770 | chunks: const [], |
| 771 | currentChunkIndex: -1, |
| 772 | ); |
| 773 | }); |
| 774 | _flutterTts!.setErrorHandler((msg) { |
| 775 | if (_isStopping || speakGeneration != _systemSpeakGeneration) return; |
| 776 | _resetSystemSeekState(); |
| 777 | state = state.copyWith( |
| 778 | isSpeaking: false, |
| 779 | isPaused: false, |
| 780 | playingContent: null, |
| 781 | error: msg.toString(), |
| 782 | clearPlayingTarget: true, |
| 783 | resetProgress: true, |
| 784 | canSeek: false, |
| 785 | chunks: const [], |
| 786 | currentChunkIndex: -1, |
| 787 | ); |
| 788 | }); |
| 789 | _flutterTts!.setCancelHandler(() { |
| 790 | if (_isStopping || speakGeneration != _systemSpeakGeneration) return; |
| 791 | }); |
| 792 | await _flutterTts!.awaitSpeakCompletion(true); |
| 793 | if (_isStopping || speakGeneration != _systemSpeakGeneration) return; |
| 794 | _systemSpeakStart = DateTime.now(); |
| 795 | _startSystemProgressTimer(speakGeneration); |
| 796 | final result = await _flutterTts!.speak(text); |
| 797 | if (result == 0 || result == -1) { |
| 798 | throw StateError('The selected TTS engine could not speak'); |
| 799 | } |
| 800 | } catch (e) { |
| 801 | if (_isStopping || speakGeneration != _systemSpeakGeneration) return; |
| 802 | _resetSystemSeekState(); |
| 803 | state = state.copyWith( |
| 804 | isSpeaking: false, |
| 805 | isPaused: false, |
| 806 | error: 'System TTS unavailable', |
| 807 | playingContent: null, |
| 808 | clearPlayingTarget: true, |
| 809 | resetProgress: true, |
| 810 | canSeek: false, |
| 811 | chunks: const [], |
| 812 | currentChunkIndex: -1, |
| 813 | ); |
| 814 | rethrow; |
| 815 | } |
| 816 | return; |
| 817 | } |
| 818 | |
| 819 | _resetProgress(); |
| 820 | _chunks = _splitText(text); |
| 821 | _currentChunkIndex = 0; |
| 822 | state = state.copyWith(chunks: _chunks, currentChunkIndex: 0); |
| 823 | _currentVoice = resolvedVoiceId; |
| 824 | _synthesizedChunks.clear(); |
| 825 | _lastEnqueuedChunk = -1; |
| 826 | _currentSessionId++; |
| 827 | _playbackSessionId = _currentSessionId; |
| 828 | |
| 829 | _isPreview = isPreview; |
| 830 | _playlistBuffer.clear(); |
| 831 | _nextPlaylistIndexToAdd = 0; |
| 832 | await _player.stop(); |
| 833 | await _player.clearAudioSources(); |
| 834 | |
| 835 | try { |
| 836 | AudioSession.instance.then((session) { |
| 837 | session.configure(const AudioSessionConfiguration.speech()); |
| 838 | }); |
| 839 | } catch (e) { |
| 840 | Log.error('Error configuring audio session: $e'); |
| 841 | } |
| 842 | |
| 843 | state = state.copyWith( |
| 844 | isSpeaking: true, |
| 845 | isPaused: false, |
| 846 | isInitializing: true, |
| 847 | error: null, |
| 848 | activeEngine: engineId, |
| 849 | playingContent: text, |
| 850 | playingMessageId: isPreview ? null : messageId, |
| 851 | playingConversationId: isPreview ? null : conversationId, |
| 852 | isPreview: isPreview, |
| 853 | spokenCharOffset: -1, |
| 854 | ); |
| 855 | |
| 856 | try { |
| 857 | await _initEngine(resolvedEngine, voiceId: resolvedVoiceId); |
| 858 | await _player.setSpeed(state.playbackSpeed); |
| 859 | state = state.copyWith( |
| 860 | isInitializing: false, |
| 861 | canSeek: false, |
| 862 | duration: _estimatedTotalDuration(), |
| 863 | ); |
| 864 | _enqueueNextChunks(); |
| 865 | } catch (e, st) { |
| 866 | if (_isPlayerDisposed) return; |
| 867 | Log.error('TTS speak error: $e\n$st'); |
| 868 | state = state.copyWith( |
| 869 | isSpeaking: false, |
| 870 | isInitializing: false, |
| 871 | error: e.toString(), |
| 872 | playingContent: null, |
| 873 | clearPlayingTarget: true, |
| 874 | ); |
| 875 | } |
| 876 | } |
| 877 | |
| 878 | Future<void> _addChunkToPlaylist(int chunkIndex, Uint8List wavBytes) async { |
| 879 | if (_isStopping || chunkIndex >= _chunks.length) return; |
| 880 | |
| 881 | final tempDir = Directory.systemTemp; |
| 882 | final tempFile = File( |
| 883 | '${tempDir.path}/tts_chunk_${_currentSessionId}_$chunkIndex.wav', |
| 884 | ); |
| 885 | await tempFile.writeAsBytes(wavBytes); |
| 886 | |
| 887 | final source = AudioSource.file( |
| 888 | tempFile.path, |
| 889 | tag: MediaItem( |
| 890 | id: 'tts_chunk_${_currentSessionId}_$chunkIndex', |
| 891 | album: _isPreview ? 'Voice Preview' : 'LocalMind TTS', |
| 892 | title: _chunks[chunkIndex], |
| 893 | artist: _currentEngine == EngineId.kitten ? 'Kitten TTS' : 'Piper TTS', |
| 894 | ), |
| 895 | ); |
| 896 | |
| 897 | _playlistBuffer[chunkIndex] = source; |
| 898 | while (_playlistBuffer.containsKey(_nextPlaylistIndexToAdd)) { |
| 899 | final indexBeingAdded = _nextPlaylistIndexToAdd; |
| 900 | final src = _playlistBuffer.remove(indexBeingAdded)!; |
| 901 | if (indexBeingAdded == 0) { |
| 902 | await _player.setAudioSources([src]); |
| 903 | } else { |
| 904 | await _player.addAudioSource(src); |
| 905 | } |
| 906 | |
| 907 | if (indexBeingAdded == 0 && !state.isPaused) { |
| 908 | await _player.setSpeed(state.playbackSpeed); |
| 909 | _player.play(); |
| 910 | } |
| 911 | |
| 912 | _nextPlaylistIndexToAdd++; |
| 913 | |
| 914 | if (_nextPlaylistIndexToAdd >= _chunks.length) { |
| 915 | state = state.copyWith(isInitializing: false, canSeek: true); |
| 916 | _saveMessageAudioCache(); |
| 917 | } else { |
| 918 | state = state.copyWith(isInitializing: false); |
| 919 | } |
| 920 | } |
| 921 | } |
| 922 | |
| 923 | void _enqueueChunk(int chunkIndex) { |
| 924 | if (_workerSendPort == null || _currentEngine == null) return; |
| 925 | if (chunkIndex >= _chunks.length) return; |
| 926 | |
| 927 | final settings = ref.read(settingsProvider); |
| 928 | final availableVoices = voicesForEngine(_currentEngine!); |
| 929 | final voiceId = |
| 930 | _currentVoice ?? |
| 931 | settings.ttsVoiceId ?? |
| 932 | (availableVoices.isNotEmpty ? availableVoices.first.id : null); |
| 933 | |
| 934 | while (_lastEnqueuedChunk < chunkIndex) { |
| 935 | _lastEnqueuedChunk++; |
| 936 | final index = _lastEnqueuedChunk; |
| 937 | Log.debug('Enqueuing synthesis for chunk $index'); |
| 938 | _workerSendPort!.send( |
| 939 | TtsSynthesizeRequest( |
| 940 | text: _chunks[index], |
| 941 | voiceId: voiceId ?? '', |
| 942 | speed: settings.ttsSpeed, |
| 943 | chunkIndex: index, |
| 944 | sessionId: _currentSessionId, |
| 945 | ), |
| 946 | ); |
| 947 | } |
| 948 | } |
| 949 | |
| 950 | Future<void> _ensurePlaylistThrough(int targetIndex) async { |
| 951 | if (targetIndex < 0 || targetIndex >= _chunks.length) return; |
| 952 | _enqueueChunk(targetIndex); |
| 953 | |
| 954 | final deadline = DateTime.now().add(const Duration(seconds: 20)); |
| 955 | while (_nextPlaylistIndexToAdd <= targetIndex && |
| 956 | DateTime.now().isBefore(deadline)) { |
| 957 | await Future.delayed(const Duration(milliseconds: 50)); |
| 958 | if (_lastEnqueuedChunk < targetIndex) { |
| 959 | _enqueueChunk(targetIndex); |
| 960 | } |
| 961 | } |
| 962 | } |
| 963 | |
| 964 | int _chunkIndexForPosition(Duration target) { |
| 965 | var accumulated = Duration.zero; |
| 966 | for (var i = 0; i < _chunks.length; i++) { |
| 967 | final chunkDuration = _chunkDurations[i] ?? _averageChunkDuration(); |
| 968 | if (accumulated + chunkDuration > target) return i; |
| 969 | accumulated += chunkDuration; |
| 970 | } |
| 971 | return _chunks.isEmpty ? 0 : _chunks.length - 1; |
| 972 | } |
| 973 | |
| 974 | Future<void> seekTo(Duration target) async { |
| 975 | if (!state.isSpeaking || |
| 976 | state.isPreview || |
| 977 | state.activeEngine == EngineId.system) { |
| 978 | return; |
| 979 | } |
| 980 | |
| 981 | state = state.copyWith(isSeekPending: true); |
| 982 | try { |
| 983 | final maxDuration = _estimatedTotalDuration(); |
| 984 | if (maxDuration <= Duration.zero) return; |
| 985 | |
| 986 | var clamped = target; |
| 987 | if (clamped.isNegative) clamped = Duration.zero; |
| 988 | if (clamped > maxDuration) clamped = maxDuration; |
| 989 | |
| 990 | final targetChunk = _chunkIndexForPosition(clamped); |
| 991 | await _ensurePlaylistThrough(targetChunk); |
| 992 | |
| 993 | var chunkStart = Duration.zero; |
| 994 | for (var i = 0; i < targetChunk; i++) { |
| 995 | chunkStart += _chunkDurations[i] ?? _averageChunkDuration(); |
| 996 | } |
| 997 | final localPosition = clamped - chunkStart; |
| 998 | |
| 999 | if (_player.currentIndex != targetChunk) { |
| 1000 | await _player.seek(Duration.zero, index: targetChunk); |
| 1001 | } |
| 1002 | if (localPosition > Duration.zero) { |
| 1003 | await _player.seek(localPosition); |
| 1004 | } else { |
| 1005 | await _player.seek(Duration.zero); |
| 1006 | } |
| 1007 | _currentChunkIndex = targetChunk; |
| 1008 | state = state.copyWith(currentChunkIndex: targetChunk); |
| 1009 | _updateChunkOffset(); |
| 1010 | _updatePlaybackProgress(); |
| 1011 | } finally { |
| 1012 | state = state.copyWith(isSeekPending: false); |
| 1013 | } |
| 1014 | } |
| 1015 | |
| 1016 | Future<void> skipForward() async { |
| 1017 | if (state.activeEngine == EngineId.system) return; |
| 1018 | if (!state.isSpeaking && !state.isPaused) return; |
| 1019 | final skip = Duration(seconds: ref.read(settingsProvider).ttsSkipSeconds); |
| 1020 | await seekTo(_computeGlobalPosition() + skip); |
| 1021 | } |
| 1022 | |
| 1023 | Future<void> skipBackward() async { |
| 1024 | if (state.activeEngine == EngineId.system) return; |
| 1025 | if (!state.isSpeaking && !state.isPaused) return; |
| 1026 | final skip = Duration(seconds: ref.read(settingsProvider).ttsSkipSeconds); |
| 1027 | await seekTo(_computeGlobalPosition() - skip); |
| 1028 | } |
| 1029 | |
| 1030 | void _enqueueNextChunks() { |
| 1031 | if (_workerSendPort == null || _currentEngine == null) return; |
| 1032 | |
| 1033 | final settings = ref.read(settingsProvider); |
| 1034 | final availableVoices = voicesForEngine(_currentEngine!); |
| 1035 | final voiceId = |
| 1036 | _currentVoice ?? |
| 1037 | settings.ttsVoiceId ?? |
| 1038 | (availableVoices.isNotEmpty ? availableVoices.first.id : null); |
| 1039 | |
| 1040 | while (_lastEnqueuedChunk < _chunks.length - 1 && |
| 1041 | _lastEnqueuedChunk < _currentChunkIndex + _lookAheadCount) { |
| 1042 | _lastEnqueuedChunk++; |
| 1043 | final chunkIndex = _lastEnqueuedChunk; |
| 1044 | final text = _chunks[chunkIndex]; |
| 1045 | |
| 1046 | Log.debug('Enqueuing synthesis for chunk $chunkIndex'); |
| 1047 | _workerSendPort!.send( |
| 1048 | TtsSynthesizeRequest( |
| 1049 | text: text, |
| 1050 | voiceId: voiceId ?? '', |
| 1051 | speed: settings.ttsSpeed, |
| 1052 | chunkIndex: chunkIndex, |
| 1053 | sessionId: _currentSessionId, |
| 1054 | ), |
| 1055 | ); |
| 1056 | } |
| 1057 | } |
| 1058 | |
| 1059 | void pause() { |
| 1060 | if (state.isPaused || !state.isSpeaking) return; |
| 1061 | if (state.activeEngine == EngineId.system) { |
| 1062 | _flutterTts?.pause(); |
| 1063 | } else { |
| 1064 | _player.pause(); |
| 1065 | } |
| 1066 | state = state.copyWith(isPaused: true); |
| 1067 | } |
| 1068 | |
| 1069 | void resume() { |
| 1070 | if (!state.isPaused) return; |
| 1071 | if (state.activeEngine == EngineId.system) { |
| 1072 | unawaited( |
| 1073 | _flutterTts?.speak(_systemFullText ?? state.playingContent ?? ''), |
| 1074 | ); |
| 1075 | } else { |
| 1076 | _player.play(); |
| 1077 | } |
| 1078 | state = state.copyWith(isPaused: false); |
| 1079 | } |
| 1080 | |
| 1081 | void togglePauseResume() { |
| 1082 | if (state.isPaused) { |
| 1083 | resume(); |
| 1084 | } else if (state.isSpeaking) { |
| 1085 | pause(); |
| 1086 | } |
| 1087 | } |
| 1088 | |
| 1089 | Future<void> stop() async { |
| 1090 | _isStopping = true; |
| 1091 | _systemSpeakGeneration++; |
| 1092 | if (_isPlayerDisposed) { |
| 1093 | _isStopping = false; |
| 1094 | return; |
| 1095 | } |
| 1096 | try { |
| 1097 | await _player.stop(); |
| 1098 | await _player.clearAudioSources(); |
| 1099 | } catch (_) {} |
| 1100 | await _flutterTts?.stop(); |
| 1101 | _cleanupSessionFiles(_currentSessionId, _chunks.length); |
| 1102 | _resetProgress(); |
| 1103 | state = state.copyWith( |
| 1104 | isSpeaking: false, |
| 1105 | isPaused: false, |
| 1106 | isInitializing: false, |
| 1107 | playingContent: null, |
| 1108 | isPreview: false, |
| 1109 | clearPlayingTarget: true, |
| 1110 | resetProgress: true, |
| 1111 | chunks: const [], |
| 1112 | currentChunkIndex: -1, |
| 1113 | canSeek: false, |
| 1114 | ); |
| 1115 | _configureAudioSessionForSpeech(); |
| 1116 | _isStopping = false; |
| 1117 | } |
| 1118 | |
| 1119 | Future<void> setPlaybackSpeed(double speed) async { |
| 1120 | if (!playbackSpeedOptions.contains(speed)) return; |
| 1121 | state = state.copyWith(playbackSpeed: speed); |
| 1122 | if (state.activeEngine == EngineId.system) { |
| 1123 | await _applySystemSpeechRate(); |
| 1124 | return; |
| 1125 | } |
| 1126 | if (!state.isSpeaking && !state.isPaused) return; |
| 1127 | try { |
| 1128 | await _player.setSpeed(speed); |
| 1129 | } catch (e) { |
| 1130 | Log.error('Failed to set playback speed: $e'); |
| 1131 | } |
| 1132 | } |
| 1133 | |
| 1134 | void cyclePlaybackSpeed() { |
| 1135 | final options = playbackSpeedOptions; |
| 1136 | final currentIndex = options.indexOf(state.playbackSpeed); |
| 1137 | final nextIndex = currentIndex < 0 |
| 1138 | ? 2 |
| 1139 | : (currentIndex + 1) % options.length; |
| 1140 | unawaited(setPlaybackSpeed(options[nextIndex])); |
| 1141 | } |
| 1142 | |
| 1143 | Future<void> _applySystemSpeechRate() async { |
| 1144 | if (_flutterTts == null) return; |
| 1145 | final settings = ref.read(settingsProvider); |
| 1146 | await _flutterTts!.setSpeechRate( |
| 1147 | (settings.ttsSpeed * state.playbackSpeed * 0.5).clamp(0.0, 1.0), |
| 1148 | ); |
| 1149 | } |
| 1150 | |
| 1151 | Future<void> _playFromCache({ |
| 1152 | required bool isPreview, |
| 1153 | required String messageId, |
| 1154 | required String? conversationId, |
| 1155 | required EngineId engineId, |
| 1156 | required String text, |
| 1157 | }) async { |
| 1158 | final cache = _messageAudioCache!; |
| 1159 | _resetProgress(); |
| 1160 | _chunks = List<String>.from(cache.chunks); |
| 1161 | _currentChunkIndex = 0; |
| 1162 | state = state.copyWith(chunks: _chunks, currentChunkIndex: 0); |
| 1163 | _currentVoice = cache.voice; |
| 1164 | _synthesizedChunks.clear(); |
| 1165 | _synthesizedChunks.addAll(cache.synthesizedChunks); |
| 1166 | _chunkDurations.clear(); |
| 1167 | _chunkDurations.addAll(cache.chunkDurations); |
| 1168 | _totalDuration = Duration.zero; |
| 1169 | for (final duration in _chunkDurations.values) { |
| 1170 | _totalDuration += duration; |
| 1171 | } |
| 1172 | _lastEnqueuedChunk = _chunks.length - 1; |
| 1173 | _currentSessionId++; |
| 1174 | _playbackSessionId = _currentSessionId; |
| 1175 | _isPreview = isPreview; |
| 1176 | _playlistBuffer.clear(); |
| 1177 | _nextPlaylistIndexToAdd = 0; |
| 1178 | await _player.stop(); |
| 1179 | |
| 1180 | state = state.copyWith( |
| 1181 | isSpeaking: true, |
| 1182 | isPaused: false, |
| 1183 | isInitializing: true, |
| 1184 | error: null, |
| 1185 | activeEngine: engineId, |
| 1186 | playingContent: text, |
| 1187 | playingMessageId: messageId, |
| 1188 | playingConversationId: conversationId, |
| 1189 | isPreview: isPreview, |
| 1190 | duration: _totalDuration, |
| 1191 | ); |
| 1192 | |
| 1193 | for (var i = 0; i < _chunks.length; i++) { |
| 1194 | final bytes = _synthesizedChunks[i]; |
| 1195 | if (bytes != null && bytes.isNotEmpty) { |
| 1196 | await _addChunkToPlaylist(i, bytes); |
| 1197 | } |
| 1198 | } |
| 1199 | |
| 1200 | state = state.copyWith(isInitializing: false, canSeek: true); |
| 1201 | if (!state.isPaused) { |
| 1202 | await _player.play(); |
| 1203 | } |
| 1204 | } |
| 1205 | |
| 1206 | void _saveMessageAudioCache() { |
| 1207 | final messageId = state.playingMessageId; |
| 1208 | final text = state.playingContent; |
| 1209 | if (state.isPreview || messageId == null || text == null) return; |
| 1210 | if (_chunks.isEmpty) return; |
| 1211 | |
| 1212 | final settings = ref.read(settingsProvider); |
| 1213 | final availableVoices = voicesForEngine( |
| 1214 | state.activeEngine ?? _currentEngine ?? EngineId.kitten, |
| 1215 | ); |
| 1216 | final resolvedVoice = |
| 1217 | _currentVoice ?? |
| 1218 | settings.ttsVoiceId ?? |
| 1219 | (availableVoices.isNotEmpty ? availableVoices.first.id : null); |
| 1220 | |
| 1221 | for (var i = 0; i < _chunks.length; i++) { |
| 1222 | final bytes = _synthesizedChunks[i]; |
| 1223 | if (bytes == null || bytes.isEmpty) return; |
| 1224 | } |
| 1225 | |
| 1226 | _messageAudioCache = _MessageAudioCache( |
| 1227 | messageId: messageId, |
| 1228 | text: text, |
| 1229 | engine: state.activeEngine ?? _currentEngine ?? EngineId.kitten, |
| 1230 | voice: resolvedVoice, |
| 1231 | chunks: List<String>.from(_chunks), |
| 1232 | synthesizedChunks: Map<int, Uint8List>.from(_synthesizedChunks), |
| 1233 | chunkDurations: Map<int, Duration>.from(_chunkDurations), |
| 1234 | ); |
| 1235 | } |
| 1236 | |
| 1237 | Future<bool> downloadCurrentAudio() async { |
| 1238 | if (state.activeEngine == EngineId.system || state.isPreview) { |
| 1239 | return false; |
| 1240 | } |
| 1241 | if (!state.isSpeaking && !state.isPaused) return false; |
| 1242 | |
| 1243 | try { |
| 1244 | final wavFiles = <Uint8List>[]; |
| 1245 | for (var i = 0; i < _chunks.length; i++) { |
| 1246 | final cached = _synthesizedChunks[i]; |
| 1247 | if (cached != null && cached.isNotEmpty) { |
| 1248 | wavFiles.add(cached); |
| 1249 | continue; |
| 1250 | } |
| 1251 | final tempFile = File( |
| 1252 | '${Directory.systemTemp.path}/tts_chunk_${_currentSessionId}_$i.wav', |
| 1253 | ); |
| 1254 | if (tempFile.existsSync()) { |
| 1255 | wavFiles.add(await tempFile.readAsBytes()); |
| 1256 | } |
| 1257 | } |
| 1258 | |
| 1259 | if (wavFiles.isEmpty) return false; |
| 1260 | |
| 1261 | final merged = mergeWavFiles(wavFiles); |
| 1262 | final timestamp = DateTime.now().millisecondsSinceEpoch; |
| 1263 | final saved = await SafeFilePicker.saveFile( |
| 1264 | dialogTitle: 'Save TTS audio', |
| 1265 | fileName: 'localmind_tts_$timestamp.wav', |
| 1266 | type: FileType.custom, |
| 1267 | allowedExtensions: const ['wav'], |
| 1268 | bytes: merged, |
| 1269 | ); |
| 1270 | return saved != null; |
| 1271 | } catch (e, st) { |
| 1272 | Log.warning('Failed to download TTS audio: $e\n$st'); |
| 1273 | return false; |
| 1274 | } |
| 1275 | } |
| 1276 | |
| 1277 | Future<void> previewVoice(Voice voice) async { |
| 1278 | await speak( |
| 1279 | ttsPreviewSample, |
| 1280 | engineId: voice.engine, |
| 1281 | voiceId: voice.id, |
| 1282 | isPreview: true, |
| 1283 | ); |
| 1284 | } |
| 1285 | } |
| 1286 | |
| 1287 | const ttsPreviewSample = "Hello, this is a preview of the selected voice"; |
| 1288 | |
| 1289 | class _MessageAudioCache { |
| 1290 | const _MessageAudioCache({ |
| 1291 | required this.messageId, |
| 1292 | required this.text, |
| 1293 | required this.engine, |
| 1294 | required this.voice, |
| 1295 | required this.chunks, |
| 1296 | required this.synthesizedChunks, |
| 1297 | required this.chunkDurations, |
| 1298 | }); |
| 1299 | |
| 1300 | final String messageId; |
| 1301 | final String text; |
| 1302 | final EngineId engine; |
| 1303 | final String? voice; |
| 1304 | final List<String> chunks; |
| 1305 | final Map<int, Uint8List> synthesizedChunks; |
| 1306 | final Map<int, Duration> chunkDurations; |
| 1307 | } |