speech to text fully drafted

lib/pages/chat/events/message.dart

@@ -145,7 +145,8 @@ class Message extends StatelessWidget {
         controller.getPangeaMessageEvent(event.eventId);
     ToolbarDisplayController? toolbarController;
     if (event.messageType == MessageTypes.Text ||
-        event.messageType == MessageTypes.Notice) {
+        event.messageType == MessageTypes.Notice ||
+        event.messageType == MessageTypes.Audio) {
       toolbarController = controller.getToolbarDisplayController(event.eventId);
     }
     // Pangea#

lib/pages/chat/events/message_content.dart

@@ -1,5 +1,6 @@
 import 'package:fluffychat/pages/chat/events/html_message.dart';
 import 'package:fluffychat/pages/chat/events/video_player.dart';
+import 'package:fluffychat/pangea/enum/message_mode_enum.dart';
 import 'package:fluffychat/pangea/models/pangea_message_event.dart';
 import 'package:fluffychat/pangea/widgets/chat/message_context_menu.dart';
 import 'package:fluffychat/pangea/widgets/chat/message_toolbar.dart';
@@ -340,7 +341,7 @@ class MessageContent extends StatelessWidget {
                               ),
                               onListen: () => toolbarController?.showToolbar(
                                 context,
-                                mode: MessageMode.play,
+                                mode: MessageMode.conversion,
                               ),
                             ),
                   enableInteractiveSelection:

lib/pangea/controllers/pangea_controller.dart

@@ -10,6 +10,7 @@ import 'package:fluffychat/pangea/controllers/local_settings.dart';
 import 'package:fluffychat/pangea/controllers/message_data_controller.dart';
 import 'package:fluffychat/pangea/controllers/my_analytics_controller.dart';
 import 'package:fluffychat/pangea/controllers/permissions_controller.dart';
+import 'package:fluffychat/pangea/controllers/speech_to_text_controller.dart';
 import 'package:fluffychat/pangea/controllers/subscription_controller.dart';
 import 'package:fluffychat/pangea/controllers/text_to_speech_controller.dart';
 import 'package:fluffychat/pangea/controllers/user_controller.dart';
@@ -47,6 +48,7 @@ class PangeaController {
   late InstructionsController instructions;
   late SubscriptionController subscriptionController;
   late TextToSpeechController textToSpeech;
+  late SpeechToTextController speechToText;
 
   ///store Services
   late PLocalStore pStoreService;
@@ -93,6 +95,7 @@ class PangeaController {
     subscriptionController = SubscriptionController(this);
     itFeedback = ITFeedbackController(this);
     textToSpeech = TextToSpeechController(this);
+    speechToText = SpeechToTextController(this);
     PAuthGaurd.pController = this;
   }
 

lib/pangea/controllers/speech_to_text_controller.dart

@@ -0,0 +1,81 @@
+import 'dart:async';
+import 'dart:convert';
+
+import 'package:fluffychat/pangea/controllers/pangea_controller.dart';
+import 'package:fluffychat/pangea/models/speech_to_text_models.dart';
+import 'package:flutter/foundation.dart';
+import 'package:http/http.dart';
+
+import '../config/environment.dart';
+import '../network/requests.dart';
+import '../network/urls.dart';
+
+// Assuming SpeechToTextRequestModel, SpeechToTextResponseModel and related models are already defined as in your provided code.
+
+class _SpeechToTextCacheItem {
+  Future<SpeechToTextResponseModel> data;
+
+  _SpeechToTextCacheItem({required this.data});
+}
+
+class SpeechToTextController {
+  static final Map<int, _SpeechToTextCacheItem> _cache = {};
+  late final PangeaController _pangeaController;
+  Timer? _cacheClearTimer;
+
+  SpeechToTextController(this._pangeaController) {
+    _initializeCacheClearing();
+  }
+
+  void _initializeCacheClearing() {
+    const duration = Duration(minutes: 15);
+    _cacheClearTimer = Timer.periodic(duration, (Timer t) => _clearCache());
+  }
+
+  void _clearCache() {
+    _cache.clear();
+  }
+
+  void dispose() {
+    _cacheClearTimer?.cancel();
+  }
+
+  Future<SpeechToTextResponseModel> get(
+      SpeechToTextRequestModel requestModel) async {
+    final int cacheKey = requestModel.hashCode;
+
+    if (_cache.containsKey(cacheKey)) {
+      return _cache[cacheKey]!.data;
+    } else {
+      final Future<SpeechToTextResponseModel> response = _fetchResponse(
+        accessToken: await _pangeaController.userController.accessToken,
+        requestModel: requestModel,
+      );
+      _cache[cacheKey] = _SpeechToTextCacheItem(data: response);
+      return response;
+    }
+  }
+
+  static Future<SpeechToTextResponseModel> _fetchResponse({
+    required String accessToken,
+    required SpeechToTextRequestModel requestModel,
+  }) async {
+    final Requests request = Requests(
+      choreoApiKey: Environment.choreoApiKey,
+      accessToken: accessToken,
+    );
+
+    final Response res = await request.post(
+      url: PApiUrls.speechToText,
+      body: requestModel.toJson(),
+    );
+
+    if (res.statusCode == 200) {
+      final Map<String, dynamic> json = jsonDecode(utf8.decode(res.bodyBytes));
+      return SpeechToTextResponseModel.fromJson(json);
+    } else {
+      debugPrint('Error converting speech to text: ${res.body}');
+      throw Exception('Failed to convert speech to text');
+    }
+  }
+}

lib/pangea/controllers/text_to_speech_controller.dart

@@ -126,25 +126,6 @@ class TextToSpeechController {
 
     return TextToSpeechResponse.fromJson(json);
   }
-  //   if (json["wave_form"] == null) {
-  //     json["wave_form"] = getWaveForm();
-  //   }
-
-  //   return TextToSpeechResponse(
-  //       audioContent: String.fromCharCodes(base64Decode(json["audio_content"])),
-  //       mediaType: json["media_type"],
-  //       durationMillis: durationMillis(json["duration_millis"]),
-  //       waveform: getWaveForm(json["audio_content"]),
-  //       );
-  // }
-
-  // static List<int> getWaveForm(audioContent) {
-  //   return [];
-  // }
-
-  // static int durationMillis(audioContent) {
-  //   return 0;
-  // }
 
   static bool isOggFile(Uint8List bytes) {
     // Check if the file has enough bytes for the header

lib/pangea/enum/audio_encoding_enum.dart

@@ -0,0 +1,56 @@
+AudioEncodingEnum mimeTypeToAudioEncoding(String mimeType) {
+  switch (mimeType) {
+    case 'audio/mpeg':
+      return AudioEncodingEnum.mp3;
+    case 'audio/mp4':
+      return AudioEncodingEnum.mp4;
+    case 'audio/ogg':
+      return AudioEncodingEnum.oggOpus;
+    default:
+      return AudioEncodingEnum.encodingUnspecified;
+  }
+}
+
+enum AudioEncodingEnum {
+  encodingUnspecified,
+  linear16,
+  flac,
+  mulaw,
+  amr,
+  amrWb,
+  oggOpus,
+  speexWithHeaderByte,
+  mp3,
+  mp4,
+  webmOpus,
+}
+
+// Utility extension to map enum values to their corresponding string value as used by the API
+extension AudioEncodingExtension on AudioEncodingEnum {
+  String get value {
+    switch (this) {
+      case AudioEncodingEnum.linear16:
+        return 'LINEAR16';
+      case AudioEncodingEnum.flac:
+        return 'FLAC';
+      case AudioEncodingEnum.mulaw:
+        return 'MULAW';
+      case AudioEncodingEnum.amr:
+        return 'AMR';
+      case AudioEncodingEnum.amrWb:
+        return 'AMR_WB';
+      case AudioEncodingEnum.oggOpus:
+        return 'OGG_OPUS';
+      case AudioEncodingEnum.speexWithHeaderByte:
+        return 'SPEEX_WITH_HEADER_BYTE';
+      case AudioEncodingEnum.mp3:
+        return 'MP3';
+      case AudioEncodingEnum.mp4:
+        return 'MP4';
+      case AudioEncodingEnum.webmOpus:
+        return 'WEBM_OPUS';
+      default:
+        return 'ENCODING_UNSPECIFIED';
+    }
+  }
+}

lib/pangea/enum/message_mode_enum.dart

@@ -0,0 +1,48 @@
+import 'package:flutter/material.dart';
+import 'package:flutter_gen/gen_l10n/l10n.dart';
+
+enum MessageMode { conversion, translation, definition }
+
+extension MessageModeExtension on MessageMode {
+  IconData icon(bool isAudioMessage) {
+    switch (this) {
+      case MessageMode.translation:
+        return Icons.g_translate;
+      case MessageMode.conversion:
+        return Icons.play_arrow;
+      //TODO change icon for audio messages
+      case MessageMode.definition:
+        return Icons.book;
+      default:
+        return Icons.error; // Icon to indicate an error or unsupported mode
+    }
+  }
+
+  String title(BuildContext context) {
+    switch (this) {
+      case MessageMode.translation:
+        return L10n.of(context)!.translations;
+      case MessageMode.conversion:
+        return L10n.of(context)!.messageAudio;
+      case MessageMode.definition:
+        return L10n.of(context)!.definitions;
+      default:
+        return L10n.of(context)!
+            .oopsSomethingWentWrong; // Title to indicate an error or unsupported mode
+    }
+  }
+
+  String tooltip(BuildContext context) {
+    switch (this) {
+      case MessageMode.translation:
+        return L10n.of(context)!.translationTooltip;
+      case MessageMode.conversion:
+        return L10n.of(context)!.audioTooltip;
+      case MessageMode.definition:
+        return L10n.of(context)!.define;
+      default:
+        return L10n.of(context)!
+            .oopsSomethingWentWrong; // Title to indicate an error or unsupported mode
+    }
+  }
+}

lib/pangea/models/pangea_message_event.dart

@@ -148,7 +148,7 @@ class PangeaMessageEvent {
       },
     );
 
-    debugPrint("eventId in getAudioGlobal $eventId");
+    debugPrint("eventId in getTextToSpeechGlobal $eventId");
     final Event? audioEvent =
         eventId != null ? await room.getEventById(eventId) : null;
 
@@ -162,10 +162,10 @@ class PangeaMessageEvent {
   //get audio for text and language
   //if no audio exists, create it
   //if audio exists, return it
-  Future<Event?> getAudioGlobal(String langCode) async {
+  Future<Event?> getTextToSpeechGlobal(String langCode) async {
     final String text = representationByLanguage(langCode)?.text ?? body;
 
-    final local = getAudioLocal(langCode, text);
+    final local = getTextToSpeechLocal(langCode, text);
 
     if (local != null) return Future.value(local);
 
@@ -223,16 +223,16 @@ class PangeaMessageEvent {
     // .timeout(
     //   Durations.long4,
     //   onTimeout: () {
-    //     debugPrint("timeout in getAudioGlobal");
+    //     debugPrint("timeout in getTextToSpeechGlobal");
     //     return null;
     //   },
     // );
 
-    debugPrint("eventId in getAudioGlobal $eventId");
+    debugPrint("eventId in getTextToSpeechGlobal $eventId");
     return eventId != null ? room.getEventById(eventId) : null;
   }
 
-  Event? getAudioLocal(String langCode, String text) {
+  Event? getTextToSpeechLocal(String langCode, String text) {
     return allAudio.firstWhereOrNull(
       (element) {
         // Safely access the transcription map

lib/pangea/models/speech_to_text_models.dart

@@ -0,0 +1,137 @@
+import 'dart:convert';
+
+import 'package:fluffychat/pangea/enum/audio_encoding_enum.dart';
+import 'package:flutter/foundation.dart';
+
+class SpeechToTextAudioConfigModel {
+  final AudioEncodingEnum encoding;
+  final int sampleRateHertz;
+  final bool enableWordConfidence;
+  final bool enableAutomaticPunctuation;
+  final String userL1;
+  final String userL2;
+
+  SpeechToTextAudioConfigModel({
+    required this.encoding,
+    required this.userL1,
+    required this.userL2,
+    this.sampleRateHertz = 16000,
+    this.enableWordConfidence = true,
+    this.enableAutomaticPunctuation = true,
+  });
+
+  Map<String, dynamic> toJson() => {
+        "encoding": encoding.value,
+        "sample_rate_hertz": sampleRateHertz,
+        "user_l1": userL1,
+        "user_l2": userL2,
+        "enable_word_confidence": enableWordConfidence,
+        "enable_automatic_punctuation": enableAutomaticPunctuation,
+      };
+}
+
+class SpeechToTextRequestModel {
+  final Uint8List audioContent;
+  final SpeechToTextAudioConfigModel config;
+
+  SpeechToTextRequestModel({
+    required this.audioContent,
+    required this.config,
+  });
+
+  Map<String, dynamic> toJson() => {
+        "audio_content": base64Encode(audioContent),
+        "config": config.toJson(),
+      };
+
+  @override
+  bool operator ==(Object other) {
+    if (identical(this, other)) return true;
+    if (other is! SpeechToTextRequestModel) return false;
+
+    return listEquals(audioContent, other.audioContent) &&
+        config == other.config;
+  }
+
+  @override
+  int get hashCode {
+    final bytesSample =
+        audioContent.length > 10 ? audioContent.sublist(0, 10) : audioContent;
+    return Object.hashAll([
+      Object.hashAll(bytesSample),
+      config.hashCode,
+    ]);
+  }
+}
+
+class WordInfo {
+  final String word;
+  final Duration? startTime;
+  final Duration? endTime;
+  final double? confidence;
+
+  WordInfo({
+    required this.word,
+    this.startTime,
+    this.endTime,
+    this.confidence,
+  });
+
+  factory WordInfo.fromJson(Map<String, dynamic> json) => WordInfo(
+        word: json['word'],
+        startTime: json['start_time'] != null
+            ? Duration(milliseconds: json['start_time'])
+            : null,
+        endTime: json['end_time'] != null
+            ? Duration(milliseconds: json['end_time'])
+            : null,
+        confidence: (json['confidence'] as num?)?.toDouble(),
+      );
+}
+
+class Transcript {
+  final String transcript;
+  final double confidence;
+  final List<WordInfo> words;
+
+  Transcript({
+    required this.transcript,
+    required this.confidence,
+    required this.words,
+  });
+
+  factory Transcript.fromJson(Map<String, dynamic> json) => Transcript(
+        transcript: json['transcript'],
+        confidence: json['confidence'].toDouble(),
+        words:
+            (json['words'] as List).map((e) => WordInfo.fromJson(e)).toList(),
+      );
+}
+
+class SpeechToTextResult {
+  final List<Transcript> transcripts;
+
+  SpeechToTextResult({required this.transcripts});
+
+  factory SpeechToTextResult.fromJson(Map<String, dynamic> json) =>
+      SpeechToTextResult(
+        transcripts: (json['transcripts'] as List)
+            .map((e) => Transcript.fromJson(e))
+            .toList(),
+      );
+}
+
+class SpeechToTextResponseModel {
+  final List<SpeechToTextResult> results;
+
+  SpeechToTextResponseModel({
+    required this.results,
+  });
+
+  factory SpeechToTextResponseModel.fromJson(Map<String, dynamic> json) =>
+      SpeechToTextResponseModel(
+        results: (json['results'] as List)
+            .map((e) => SpeechToTextResult.fromJson(e))
+            .toList(),
+      );
+}

lib/pangea/network/urls.dart

@@ -51,6 +51,7 @@ class PApiUrls {
   static String subseqStep = "/it_step";
 
   static String textToSpeech = "${Environment.choreoApi}/text_to_speech";
+  static String speechToText = "${Environment.choreoApi}/speech_to_text";
 
   ///-------------------------------- revenue cat --------------------------
   static String rcApiV1 = "https://api.revenuecat.com/v1";

lib/pangea/repo/text_to_speech_repo.dart

@@ -1,66 +0,0 @@
-// import 'dart:async';
-// import 'dart:convert';
-
-// import 'package:fluffychat/pangea/config/environment.dart';
-// import 'package:fluffychat/pangea/constants/model_keys.dart';
-// import 'package:fluffychat/pangea/network/urls.dart';
-// import 'package:http/http.dart';
-
-// import '../network/requests.dart';
-
-// class TextToSpeechRequest {
-//   String text;
-//   String langCode;
-
-//   TextToSpeechRequest({required this.text, required this.langCode});
-
-//   Map<String, dynamic> toJson() => {
-//         ModelKey.text: text,
-//         ModelKey.langCode: langCode,
-//       };
-// }
-
-// class TextToSpeechResponse {
-//   String audioContent;
-//   String mediaType;
-//   int durationMillis;
-//   List<int> waveform;
-
-//   TextToSpeechResponse({
-//     required this.audioContent,
-//     required this.mediaType,
-//     required this.durationMillis,
-//     required this.waveform,
-//   });
-
-//   factory TextToSpeechResponse.fromJson(
-//     Map<String, dynamic> json,
-//   ) =>
-//       TextToSpeechResponse(
-//         audioContent: json["audio_content"],
-//         mediaType: json["media_type"],
-//         durationMillis: json["duration_millis"],
-//         waveform: List<int>.from(json["wave_form"]),
-//       );
-// }
-
-// class TextToSpeechService {
-//   static Future<TextToSpeechResponse> get({
-//     required String accessToken,
-//     required TextToSpeechRequest params,
-//   }) async {
-//     final Requests request = Requests(
-//       choreoApiKey: Environment.choreoApiKey,
-//       accessToken: accessToken,
-//     );
-
-//     final Response res = await request.post(
-//       url: PApiUrls.textToSpeech,
-//       body: params.toJson(),
-//     );
-
-//     final Map<String, dynamic> json = jsonDecode(res.body);
-
-//     return TextToSpeechResponse.fromJson(json);
-//   }
-// }

lib/pangea/widgets/chat/message_audio_card.dart

@@ -33,7 +33,7 @@ class MessageAudioCardState extends State<MessageAudioCard> {
           widget.messageEvent.representationByLanguage(langCode)?.text;
       if (text != null) {
         final Event? localEvent =
-            widget.messageEvent.getAudioLocal(langCode, text);
+            widget.messageEvent.getTextToSpeechLocal(langCode, text);
         if (localEvent != null) {
           localAudioEvent = localEvent;
           if (mounted) setState(() => _isLoading = false);

lib/pangea/widgets/chat/message_speech_to_text_card.dart

@@ -0,0 +1,187 @@
+import 'dart:developer';
+
+import 'package:fluffychat/pangea/enum/audio_encoding_enum.dart';
+import 'package:fluffychat/pangea/models/message_data_models.dart';
+import 'package:fluffychat/pangea/models/pangea_message_event.dart';
+import 'package:fluffychat/pangea/models/speech_to_text_models.dart';
+import 'package:fluffychat/pangea/utils/bot_style.dart';
+import 'package:fluffychat/pangea/utils/error_handler.dart';
+import 'package:fluffychat/utils/localized_exception_extension.dart';
+import 'package:fluffychat/widgets/matrix.dart';
+import 'package:flutter/foundation.dart';
+import 'package:flutter/material.dart';
+import 'package:matrix/matrix.dart';
+
+class MessageSpeechToTextCard extends StatefulWidget {
+  final PangeaMessageEvent messageEvent;
+
+  const MessageSpeechToTextCard({
+    super.key,
+    required this.messageEvent,
+  });
+
+  @override
+  MessageSpeechToTextCardState createState() => MessageSpeechToTextCardState();
+}
+
+enum AudioFileStatus { notDownloaded, downloading, downloaded }
+
+class MessageSpeechToTextCardState extends State<MessageSpeechToTextCard> {
+  PangeaRepresentation? repEvent;
+  String? transcription;
+  bool _fetchingTranscription = true;
+  AudioFileStatus status = AudioFileStatus.notDownloaded;
+  MatrixFile? matrixFile;
+  // File? audioFile;
+
+  String? get l1Code =>
+      MatrixState.pangeaController.languageController.activeL1Code(
+        roomID: widget.messageEvent.room.id,
+      );
+  String? get l2Code =>
+      MatrixState.pangeaController.languageController.activeL2Code(
+        roomID: widget.messageEvent.room.id,
+      );
+
+  // get transcription from local events
+  Future<String> getLocalTranscription() async {
+    return "This is a dummy transcription";
+  }
+
+  // This code is duplicated from audio_player.dart. Is there some way to reuse that code?
+  Future<void> _downloadAction() async {
+    // #Pangea
+    // if (status != AudioFileStatus.notDownloaded) return;
+    if (status != AudioFileStatus.notDownloaded) {
+      return;
+    }
+    // Pangea#
+    setState(() => status = AudioFileStatus.downloading);
+    try {
+      // #Pangea
+      // final matrixFile = await widget.event.downloadAndDecryptAttachment();
+      final matrixFile =
+          await widget.messageEvent.event.downloadAndDecryptAttachment();
+      // Pangea#
+      // File? file;
+
+      // TODO: Test on mobile and see if we need this case
+      // if (!kIsWeb) {
+      //   final tempDir = await getTemporaryDirectory();
+      //   final fileName = Uri.encodeComponent(
+      //     // #Pangea
+      //     // widget.event.attachmentOrThumbnailMxcUrl()!.pathSegments.last,
+      //     widget.messageEvent.event
+      //         .attachmentOrThumbnailMxcUrl()!
+      //         .pathSegments
+      //         .last,
+      //     // Pangea#
+      //   );
+      //   file = File('${tempDir.path}/${fileName}_${matrixFile.name}');
+      //   await file.writeAsBytes(matrixFile.bytes);
+      // }
+
+      // audioFile = file;
+      this.matrixFile = matrixFile;
+      status = AudioFileStatus.downloaded;
+    } catch (e, s) {
+      Logs().v('Could not download audio file', e, s);
+      ScaffoldMessenger.of(context).showSnackBar(
+        SnackBar(
+          content: Text(e.toLocalizedString(context)),
+        ),
+      );
+    }
+  }
+
+  AudioEncodingEnum? get encoding {
+    if (matrixFile == null) return null;
+    return mimeTypeToAudioEncoding(matrixFile!.mimeType);
+  }
+
+  // call API to transcribe audio
+  Future<String?> transcribeAudio() async {
+    await _downloadAction();
+
+    final localmatrixFile = matrixFile;
+    final info = matrixFile?.info;
+
+    if (matrixFile == null) {
+      debugger(when: kDebugMode);
+      ErrorHandler.logError(
+        e: 'Audio file or matrix file is null ${widget.messageEvent.eventId}',
+        s: StackTrace.current,
+        data: widget.messageEvent.event.content,
+      );
+      return null;
+    }
+
+    if (l1Code == null || l2Code == null) {
+      debugger(when: kDebugMode);
+      ErrorHandler.logError(
+        e: 'Language codes are null ${widget.messageEvent.eventId}',
+        s: StackTrace.current,
+        data: widget.messageEvent.event.content,
+      );
+      return null;
+    }
+
+    final SpeechToTextResponseModel response =
+        await MatrixState.pangeaController.speechToText.get(
+      SpeechToTextRequestModel(
+        audioContent: matrixFile!.bytes,
+        config: SpeechToTextAudioConfigModel(
+          encoding: encoding ?? AudioEncodingEnum.encodingUnspecified,
+          //this is the default in the RecordConfig in record package
+          sampleRateHertz: 44100,
+          userL1: l1Code!,
+          userL2: l2Code!,
+        ),
+      ),
+    );
+    return response.results.first.transcripts.first.transcript;
+  }
+
+  // look for transcription in message event
+  // if not found, call API to transcribe audio
+  Future<void> loadTranscription() async {
+    // transcription ??= await getLocalTranscription();
+    transcription ??= await transcribeAudio();
+    setState(() => _fetchingTranscription = false);
+  }
+
+  @override
+  void initState() {
+    super.initState();
+    loadTranscription();
+  }
+
+  @override
+  Widget build(BuildContext context) {
+    // if (!_fetchingTranscription && repEvent == null && transcription == null) {
+    //   return const CardErrorWidget();
+    // }
+
+    return Padding(
+      padding: const EdgeInsets.all(8),
+      child: _fetchingTranscription
+          ? SizedBox(
+              height: 14,
+              width: 14,
+              child: CircularProgressIndicator(
+                strokeWidth: 2.0,
+                color: Theme.of(context).colorScheme.primary,
+              ),
+            )
+          : transcription != null
+              ? Text(
+                  transcription!,
+                  style: BotStyle.text(context),
+                )
+              : Text(
+                  repEvent!.text,
+                  style: BotStyle.text(context),
+                ),
+    );
+  }
+}

lib/pangea/widgets/chat/message_toolbar.dart

@@ -1,25 +1,27 @@
 import 'dart:async';
+import 'dart:developer';
 
 import 'package:fluffychat/config/themes.dart';
 import 'package:fluffychat/pages/chat/chat.dart';
 import 'package:fluffychat/pangea/constants/local.key.dart';
+import 'package:fluffychat/pangea/enum/message_mode_enum.dart';
 import 'package:fluffychat/pangea/models/pangea_message_event.dart';
 import 'package:fluffychat/pangea/utils/any_state_holder.dart';
 import 'package:fluffychat/pangea/utils/error_handler.dart';
 import 'package:fluffychat/pangea/utils/overlay.dart';
 import 'package:fluffychat/pangea/widgets/chat/message_audio_card.dart';
+import 'package:fluffychat/pangea/widgets/chat/message_speech_to_text_card.dart';
 import 'package:fluffychat/pangea/widgets/chat/message_text_selection.dart';
 import 'package:fluffychat/pangea/widgets/chat/message_translation_card.dart';
 import 'package:fluffychat/pangea/widgets/chat/message_unsubscribed_card.dart';
 import 'package:fluffychat/pangea/widgets/chat/overlay_message.dart';
 import 'package:fluffychat/pangea/widgets/igc/word_data_card.dart';
 import 'package:fluffychat/widgets/matrix.dart';
+import 'package:flutter/foundation.dart';
 import 'package:flutter/material.dart';
 import 'package:flutter_gen/gen_l10n/l10n.dart';
 import 'package:matrix/matrix.dart';
 
-enum MessageMode { translation, play, definition }
-
 class ToolbarDisplayController {
   final PangeaMessageEvent pangeaMessageEvent;
   final String targetId;
@@ -90,6 +92,7 @@ class ToolbarDisplayController {
           ],
         );
       } catch (err) {
+        debugger(when: kDebugMode);
         ErrorHandler.logError(e: err, s: StackTrace.current);
         return;
       }
@@ -147,53 +150,12 @@ class MessageToolbar extends StatefulWidget {
 }
 
 class MessageToolbarState extends State<MessageToolbar> {
-  Widget? child;
+  Widget? toolbarContent;
   MessageMode? currentMode;
   bool updatingMode = false;
   late StreamSubscription<String?> selectionStream;
   late StreamSubscription<MessageMode> toolbarModeStream;
 
-  IconData getIconData(MessageMode mode) {
-    switch (mode) {
-      case MessageMode.translation:
-        return Icons.g_translate;
-      case MessageMode.play:
-        return Icons.play_arrow;
-      case MessageMode.definition:
-        return Icons.book;
-      default:
-        return Icons.error; // Icon to indicate an error or unsupported mode
-    }
-  }
-
-  String getModeTitle(MessageMode mode) {
-    switch (mode) {
-      case MessageMode.translation:
-        return L10n.of(context)!.translations;
-      case MessageMode.play:
-        return L10n.of(context)!.messageAudio;
-      case MessageMode.definition:
-        return L10n.of(context)!.definitions;
-      default:
-        return L10n.of(context)!
-            .oopsSomethingWentWrong; // Title to indicate an error or unsupported mode
-    }
-  }
-
-  String getModeTooltip(MessageMode mode) {
-    switch (mode) {
-      case MessageMode.translation:
-        return L10n.of(context)!.translationTooltip;
-      case MessageMode.play:
-        return L10n.of(context)!.audioTooltip;
-      case MessageMode.definition:
-        return L10n.of(context)!.define;
-      default:
-        return L10n.of(context)!
-            .oopsSomethingWentWrong; // Title to indicate an error or unsupported mode
-    }
-  }
-
   void updateMode(MessageMode newMode) {
     if (updatingMode) return;
     debugPrint("updating toolbar mode");
@@ -204,8 +166,8 @@ class MessageToolbarState extends State<MessageToolbar> {
       updatingMode = true;
     });
     if (!subscribed) {
-      child = MessageUnsubscribedCard(
-        languageTool: getModeTitle(newMode),
+      toolbarContent = MessageUnsubscribedCard(
+        languageTool: newMode.title(context),
         mode: newMode,
         toolbarModeStream: widget.toolbarModeStream,
       );
@@ -214,8 +176,8 @@ class MessageToolbarState extends State<MessageToolbar> {
         case MessageMode.translation:
           showTranslation();
           break;
-        case MessageMode.play:
-          playAudio();
+        case MessageMode.conversion:
+          showConversion();
           break;
         case MessageMode.definition:
           showDefinition();
@@ -231,28 +193,35 @@ class MessageToolbarState extends State<MessageToolbar> {
 
   void showTranslation() {
     debugPrint("show translation");
-    child = MessageTranslationCard(
+    toolbarContent = MessageTranslationCard(
       messageEvent: widget.pangeaMessageEvent,
       immersionMode: widget.immersionMode,
       selection: widget.textSelection,
     );
   }
 
-  void playAudio() {
-    debugPrint("play audio");
-    child = MessageAudioCard(
-      messageEvent: widget.pangeaMessageEvent,
-    );
+  void showConversion() {
+    debugPrint("show conversion");
+    if (isAudioMessage) {
+      debugPrint("is audio message");
+      toolbarContent = MessageSpeechToTextCard(
+        messageEvent: widget.pangeaMessageEvent,
+      );
+    } else {
+      toolbarContent = MessageAudioCard(
+        messageEvent: widget.pangeaMessageEvent,
+      );
+    }
   }
 
   void showDefinition() {
     if (widget.textSelection.selectedText == null ||
         widget.textSelection.selectedText!.isEmpty) {
-      child = const SelectToDefine();
+      toolbarContent = const SelectToDefine();
       return;
     }
 
-    child = WordDataCard(
+    toolbarContent = WordDataCard(
       word: widget.textSelection.selectedText!,
       wordLang: widget.pangeaMessageEvent.messageDisplayLangCode,
       fullText: widget.textSelection.messageText,
@@ -262,6 +231,10 @@ class MessageToolbarState extends State<MessageToolbar> {
     );
   }
 
+  bool get isAudioMessage {
+    return widget.pangeaMessageEvent.event.messageType == MessageTypes.Audio;
+  }
+
   void showImage() {}
 
   void spellCheck() {}
@@ -286,7 +259,7 @@ class MessageToolbarState extends State<MessageToolbar> {
           ) ??
           true;
       autoplay
-          ? updateMode(MessageMode.play)
+          ? updateMode(MessageMode.conversion)
           : updateMode(MessageMode.translation);
     });
 
@@ -339,8 +312,8 @@ class MessageToolbarState extends State<MessageToolbar> {
                   duration: FluffyThemes.animationDuration,
                   child: Column(
                     children: [
-                      child ?? const SizedBox(),
-                      SizedBox(height: child == null ? 0 : 20),
+                      toolbarContent ?? const SizedBox(),
+                      SizedBox(height: toolbarContent == null ? 0 : 20),
                     ],
                   ),
                 ),
@@ -349,10 +322,13 @@ class MessageToolbarState extends State<MessageToolbar> {
             Row(
               mainAxisSize: MainAxisSize.min,
               children: MessageMode.values.map((mode) {
+                    if (mode == MessageMode.definition && isAudioMessage) {
+                      return const SizedBox.shrink();
+                    }
                     return Tooltip(
-                      message: getModeTooltip(mode),
+                      message: mode.tooltip(context),
                       child: IconButton(
-                        icon: Icon(getIconData(mode)),
+                        icon: Icon(mode.icon(isAudioMessage)),
                         color: currentMode == mode
                             ? Theme.of(context).colorScheme.primary
                             : null,

lib/pangea/widgets/chat/message_unsubscribed_card.dart

@@ -2,11 +2,12 @@ import 'dart:async';
 
 import 'package:fluffychat/config/app_config.dart';
 import 'package:fluffychat/pangea/utils/bot_style.dart';
-import 'package:fluffychat/pangea/widgets/chat/message_toolbar.dart';
 import 'package:fluffychat/widgets/matrix.dart';
 import 'package:flutter/material.dart';
 import 'package:flutter_gen/gen_l10n/l10n.dart';
 
+import '../../enum/message_mode_enum.dart';
+
 class MessageUnsubscribedCard extends StatelessWidget {
   final String languageTool;
   final MessageMode mode;

lib/pangea/widgets/chat/text_to_speech_button.dart

@@ -50,7 +50,7 @@ class _TextToSpeechButtonState extends State<TextToSpeechButton> {
 
   Event? get localAudioEvent =>
       langCode != null && text != null && text!.isNotEmpty
-          ? _pangeaMessageEvent.getAudioLocal(langCode!, text!)
+          ? _pangeaMessageEvent.getTextToSpeechLocal(langCode!, text!)
           : null;
 
   String? get langCode =>
@@ -69,7 +69,7 @@ class _TextToSpeechButtonState extends State<TextToSpeechButton> {
       if (langCode == null || langCode!.isEmpty) return;
 
       setState(() => _isLoading = true);
-      await _pangeaMessageEvent.getAudioGlobal(langCode!);
+      await _pangeaMessageEvent.getTextToSpeechGlobal(langCode!);
       setState(() => _isLoading = false);
     } catch (e) {
       setState(() => _isLoading = false);

lib/pangea/widgets/igc/pangea_rich_text.dart

@@ -14,6 +14,7 @@ import 'package:fluffychat/widgets/matrix.dart';
 import 'package:flutter/foundation.dart';
 import 'package:flutter/material.dart';
 
+import '../../enum/message_mode_enum.dart';
 import '../../models/pangea_match_model.dart';
 
 class PangeaRichText extends StatefulWidget {
@@ -157,7 +158,7 @@ class PangeaRichTextState extends State<PangeaRichText> {
                   ),
                   onListen: () => widget.toolbarController?.showToolbar(
                     context,
-                    mode: MessageMode.play,
+                    mode: MessageMode.conversion,
                   ),
                 ),
       TextSpan(