rollback tts

app.py

@@ -28,15 +28,15 @@ def create_vc_fn(tgt_sr, net_g, vc, if_f0, file_index, file_big_npy):
         index_rate,
         tts_mode,
         tts_text,
-        tts_speaker
+        tts_voice
     ):
         try:
             if tts_mode:
                 if len(tts_text) > 600 and limitation:
                     return "Text is too long", None
-                if tts_text is None or tts_speaker is None:
+                if tts_text is None or tts_voice is None:
                     return "You need to enter text and select a voice", None
-                asyncio.run(edge_tts.Communicate(tts_text, tts_speaker).save("tts.mp3"))
+                asyncio.run(edge_tts.Communicate(tts_text, "-".join(tts_voice.split('-')[:-1])).save("tts.mp3"))
                 audio, sr = librosa.load("tts.mp3", sr=16000, mono=True)
             else:
                 if args.files:
@@ -106,8 +106,8 @@ if __name__ == '__main__':
     args, unknown = parser.parse_known_args()
     load_hubert()
     models = []
-    tts_speakers_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
-    speaker = tts_speakers_list[tts_speaker]['ShortName']
+    tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
+    voices = [f"{v['ShortName']}-{v['Gender']}" for v in tts_voice_list]
     with open("weights/model_info.json", "r", encoding="utf-8") as f:
         models_info = json.load(f)
     for name, info in models_info.items():
@@ -175,21 +175,11 @@ if __name__ == '__main__':
                             )
                             tts_mode = gr.Checkbox(label="tts (use edge-tts as input)", value=False)
                             tts_text = gr.Textbox(visible=False,label="TTS text (600 words limitation)" if limitation else "TTS text")
-                            tts_speaker = gr.Dropdown(
-                                [
-                                    '%s (%s)' % (
-                                        s['FriendlyName'],
-                                        s['Gender']
-                                    )
-                                    for s in tts_speakers_list
-                                ],
-                                label='TTS speaker',
-                                type='index'
-                            )
+                            tts_voice = gr.Dropdown(label="Edge-tts speaker", choices=voices, visible=False, allow_custom_value=False, value="en-US-AnaNeural-Female")
                             vc_submit = gr.Button("Generate", variant="primary")
                         with gr.Column():
                             vc_output1 = gr.Textbox(label="Output Message")
                             vc_output2 = gr.Audio(label="Output Audio")
-                vc_submit.click(vc_fn, [vc_input, vc_transpose, vc_f0method, vc_index_ratio, tts_mode, tts_text, tts_speaker], [vc_output1, vc_output2])
-                tts_mode.change(change_to_tts_mode, [tts_mode], [vc_input, tts_text, tts_speaker])
+                vc_submit.click(vc_fn, [vc_input, vc_transpose, vc_f0method, vc_index_ratio, tts_mode, tts_text, tts_voice], [vc_output1, vc_output2])
+                tts_mode.change(change_to_tts_mode, [tts_mode], [vc_input, tts_text, tts_voice])
         app.queue(concurrency_count=1, max_size=20, api_open=args.api).launch(share=args.share)