Add --default-voice option to mimic3-server and split voice into language/name

mimic3_http/app.py

@@ -36,6 +36,7 @@ from quart import (
 from swagger_ui import api_doc
 
 from mimic3_tts import DEFAULT_VOICE, Mimic3Settings, Mimic3TextToSpeechSystem
+from mimic3_tts.utils import LANG_NAMES
 
 from ._resources import _DIR, _PACKAGE
 from .args import _MISSING
@@ -157,6 +158,7 @@ def get_app(args: argparse.Namespace, request_queue: Queue, temp_dir: str):
             "index.html",
             show_openapi=show_openapi,
             max_text_length=args.max_text_length,
+            default_voice=args.default_voice,
         )
 
     @app.route("/api/tts", methods=["GET", "POST"])
@@ -225,9 +227,26 @@ def get_app(args: argparse.Namespace, request_queue: Queue, temp_dir: str):
 
     @app.route("/api/voices", methods=["GET"])
     async def api_voices():
-        voices_dict = {v.key: v for v in _MIMIC3.get_voices()}
+        voices_by_key = {v.key: v for v in _MIMIC3.get_voices()}
-        voices = sorted(voices_dict.values(), key=lambda v: v.key)
+        sorted_voices = sorted(voices_by_key.values(), key=lambda v: v.key)
-        return jsonify([dataclasses.asdict(v) for v in voices])
+        voice_dicts = [dataclasses.asdict(v) for v in sorted_voices]
+
+        # Add more fields to voices
+        for voice_dict in voice_dicts:
+            voice_lang = voice_dict["language"]
+            lang_name = LANG_NAMES.get(voice_lang, voice_lang)
+
+            if isinstance(lang_name, str):
+                # Native and English language name are the same
+                native_lang, english_lang = lang_name, lang_name
+            else:
+                # Native and English language name are different
+                native_lang, english_lang = lang_name
+
+            voice_dict["language_native"] = native_lang
+            voice_dict["language_english"] = english_lang
+
+        return jsonify(voice_dicts)
 
     @app.route("/process", methods=["GET", "POST"])
     async def api_process():

mimic3_http/args.py

@@ -86,6 +86,10 @@ def get_args(argv=None) -> argparse.Namespace:
         type=int,
         help="Maximum length of input text to process (default: no limit)",
     )
+    parser.add_argument(
+        "--default-voice",
+        help="Default voice key to select in web interface",
+    )
     parser.add_argument(
         "--debug", action="store_true", help="Print DEBUG messages to console"
     )

mimic3_http/templates/index.html

@@ -87,8 +87,13 @@
             </div>
             <div class="row mt-3">
                 <div class="col-auto">
-                    <label for="voice-list" title="Voice name">Voice:</label>
+                    <label for="voice-language" title="Voice language">Language:</label>
-                    <select id="voice-list" name="voices">
+                    <select id="voice-language" name="voice-language">
+                    </select>
+                </div>
+                <div class="col-auto">
+                    <label for="voice-name" title="Voice name">Name:</label>
+                    <select id="voice-name" name="voice-name">
                     </select>
                 </div>
                 <div class="col-auto">
@@ -246,8 +251,8 @@
          q('#speak-button').addEventListener('click', do_tts)
 
          async function synthesize(text) {
-             var voiceList = q('#voice-list')
+             var voiceName = q('#voice-name')
-             var voice = voiceList.options[voiceList.selectedIndex].value
+             var voice = voiceName.options[voiceName.selectedIndex].value
 
              var noiseScale = q('#noise-scale').value || '0.667'
              var noiseW = q('#noise-w').value || '0.8'
@@ -289,8 +294,52 @@
              }
          }
 
-         function voiceChanged() {
+         function langChanged(indexToSelect) {
-             var voiceList = q('#voice-list')
+             // Called when voice language is changed
+
+             // If set, select a specific language by index
+             indexToSelect = indexToSelect === undefined ? -1 : indexToSelect
+             
+             var voiceLang = q('#voice-language')
+
+             // Reset names
+             var voiceName = q('#voice-name')
+             for (var i = voiceName.options.length - 1; i >= 0; i--) {
+                 voiceName.options[i].remove()
+             }
+
+             if (indexToSelect >= 0) {
+                 // Select specific language
+                 voiceLang.selectedIndex = indexToSelect
+             }
+
+             var selectedLang = voiceLang.options[voiceLang.selectedIndex].value
+             var nameIndexToSelect = -1
+
+             Object.values(voicesInfo).forEach(function(voice) {
+                 if (voice.language == selectedLang) {
+                     voiceName.insertAdjacentHTML(
+                         'beforeend', '<option value="' + voice.key + '">' + voice.name + '</option>'
+                     )
+
+                     if ((indexToSelect >= 0) && (voice.key == '{{ default_voice }}')) {
+                         // Record voice name index to select
+                         nameIndexToSelect = voiceName.options.length - 1                         
+                     }
+                 }
+             })
+                 
+             // Trigger voice name change
+             nameChanged(nameIndexToSelect)
+         }
+
+         function nameChanged(indexToSelect) {
+             // Called when voice name is changed
+
+             // If set, select a specific voice by index
+             indexToSelect = indexToSelect === undefined ? -1 : indexToSelect
+
+             var voiceName = q('#voice-name')
 
              // Reset audio
              q('#audio-message').hidden = true
@@ -304,7 +353,12 @@
                  speakerList.options[i].remove()
              }
 
-             var voiceKey = voiceList.options[voiceList.selectedIndex].value
+             if (indexToSelect >= 0) {
+                 // Select a specific voice by index
+                 voiceName.selectedIndex = indexToSelect
+             }
+
+             var voiceKey = voiceName.options[voiceName.selectedIndex].value
              var voice = voicesInfo[voiceKey]
 
              if (voice.speakers && voice.speakers.length > 0) {
@@ -329,17 +383,22 @@
              }
          }
 
-         q('#voice-list').addEventListener('change', voiceChanged)
-
          function loadVoices() {
              voicesInfo = {}
 
              // Remove previous voices
-             var voiceList = q('#voice-list')
+             var voiceLang = q('#voice-language')
-             for (var i = voiceList.options.length - 1; i >= 0; i--) {
+             for (var i = voiceLang.options.length - 1; i >= 0; i--) {
-                 voiceList.options[i].remove()
+                 voiceLang.options[i].remove()
              }
 
+             var voiceName = q('#voice-name')
+             for (var i = voiceName.options.length - 1; i >= 0; i--) {
+                 voiceName.options[i].remove()
+             }
+
+             var langs = new Set();
+
              fetch('api/voices')
                  .then(function(res) {
                      if (!res.ok) throw Error(res.statusText)
@@ -352,12 +411,19 @@
 
                      voices.forEach(function(voice) {
                          voicesInfo[voice.key] = voice
-                         voiceList.insertAdjacentHTML(
+                         if (!langs.has(voice.language)) {
-                             'beforeend', '<option value="' + voice.key + '">' + voice.language + '/' + voice.name + '</option>'
+                             voiceLang.insertAdjacentHTML(
-                         )
+                                 'beforeend', '<option value="' + voice.language + '" title="' + voice.language_english + '">' + voice.language_native + '</option>'
+                             )
+                             langs.add(voice.language)                            
+                         }
+
+                         if (voice.key == '{{ default_voice }}') {
+                             indexToSelect = voiceLang.options.length - 1
+                         }
                      })
 
-                     voiceChanged()
+                     langChanged(indexToSelect)
                  }).catch(function(err) {
                      q('#message').textContent = 'Error: ' + err.message
                      q('#speak-button').disabled = false
@@ -366,6 +432,10 @@
 
          window.addEventListener('load', function() {
              loadVoices()
+
+             q('#voice-language').addEventListener('change', langChanged)
+             q('#voice-name').addEventListener('change', nameChanged)
+
          })
         </script>
 

mimic3_tts/utils.py

@@ -28,8 +28,8 @@ LANG_NAMES = {
     "af_ZA": "Afrikaans",
     "da_DK": ("Dansk", "Danish"),
     "de_DE": ("Deutsch", "German"),
-    "en_UK": "English",
+    "en_UK": "English (UK)",
-    "en_US": "English",
+    "en_US": "English (US)",
     "el_GR": ("Ελληνικά", "Greek"),
     "es_ES": ("Español", "Spanish"),
     "fa": ("فارسی", "Persian"),