update

app.py

@@ -1,5 +1,6 @@
 import gradio as gr
 import os, torch, io
+import sys
 #os.system('python -m unidic download')
 from melo.api import TTS
 speed = 1.0
@@ -15,6 +16,8 @@ else:
 languages = ["EN", "ES", "FR", "ZH", "JP", "KR"]
 en = ["EN-Default", "EN-US", "EN-BR", "EN_INDIA", "EN-AU"]
 
+LANG = sys.argv[1]
+
 #model = TTS(language='EN', device=device)
 def synthesize(language, speaker, text, speed=1.0, progress=gr.Progress()):
     model = TTS(language=language, device=device)
@@ -23,19 +26,22 @@ def synthesize(language, speaker, text, speed=1.0, progress=gr.Progress()):
     model.tts_to_file(text, speaker_ids[speaker], bio, speed=speed, pbar=progress.tqdm, format='wav')
     return bio.getvalue()
 
-def lang(language):
-    if language == "EN":
-        return gr.update(choices=en, value="EN-Default")
-    else:
-        return gr.update(choices=[language], value=language)
+#def lang(language):
+#    if language == "EN":
+#        return gr.update(choices=en, value="EN-Default")
+#    else:
+#        return gr.update(choices=[language], value=language)
 with gr.Blocks() as demo:
     gr.Markdown('# MeloTTS\n\nAn unofficial demo of [MeloTTS](https://github.com/myshell-ai/MeloTTS) from MyShell AI. MeloTTS is a permissively licensed (MIT) SOTA multi-speaker TTS model.\n\nI am not affiliated with MyShell AI in any way.\n\nThis demo currently only supports English, but the model itself supports other languages.')
     with gr.Group():
-        language = gr.Dropdown(languages, interactive=True, value='EN', label='Language')
-        speaker = gr.Dropdown(en, interactive=True, value='EN-Default', label='Speaker')
+#        language = gr.Dropdown(languages, interactive=True, value='EN', label='Language')
+        if LANG == "EN":
+            speaker = gr.Dropdown(en, interactive=True, value='EN-Default', label='Speaker')
+        else:
+            speaker = gr.Dropdown([LANG], interactive=True, value='EN-Default', label='Speaker')
         speed = gr.Slider(label='Speed', minimum=0.1, maximum=10.0, value=1.0, interactive=True, step=0.1)
         text = gr.Textbox(label="Text to speak", value='The field of text to speech has seen rapid development recently')
-    language.change(fn=lang, inputs=[language], outputs=[speaker])
+#    language.change(fn=lang, inputs=[language], outputs=[speaker])
     btn = gr.Button('Synthesize', variant='primary')
     aud = gr.Audio(interactive=False)
     btn.click(synthesize, inputs=[language, speaker, text, speed], outputs=[aud])

melo/text/japanese.py

@@ -11,7 +11,8 @@ punctuation = ["!", "?", "…", ",", ".", "'", "-"]
 try:
     import MeCab
 except ImportError as e:
-    raise ImportError("Japanese requires mecab-python3 and unidic-lite.") from e
+    print("")
+#    raise ImportError("Japanese requires mecab-python3 and unidic-lite.") from e
 from num2words import num2words
 
 _CONVRULES = [
@@ -364,7 +365,11 @@ def hira2kata(text: str) -> str:
 
 _SYMBOL_TOKENS = set(list("・、。？！"))
 _NO_YOMI_TOKENS = set(list("「」『』―（）［］[]"))
-_TAGGER = MeCab.Tagger()
+
+try:
+    _TAGGER = MeCab.Tagger()
+except:
+    print("")
 
 
 def text2kata(text: str) -> str:
@@ -644,4 +649,4 @@ if __name__ == "__main__":
 #     conv = kakasi.getConverter()
 #     katakana_text = conv.do('ええ、僕はおきなと申します。こちらの小さいわらべは杏子。ご挨拶が遅れてしまいすみません。あなたの名は?')  # Replace with your Chinese text
 
-#     print(katakana_text)  # Output: ニーハオセカイ
+#     print(katakana_text)  # Output: ニーハオセカイ