Spaces:
Runtime error
Runtime error
Commit
•
e383d75
1
Parent(s):
96fddc0
update
Browse files- app.py +14 -8
- melo/text/japanese.py +8 -3
app.py
CHANGED
@@ -1,5 +1,6 @@
|
|
1 |
import gradio as gr
|
2 |
import os, torch, io
|
|
|
3 |
#os.system('python -m unidic download')
|
4 |
from melo.api import TTS
|
5 |
speed = 1.0
|
@@ -15,6 +16,8 @@ else:
|
|
15 |
languages = ["EN", "ES", "FR", "ZH", "JP", "KR"]
|
16 |
en = ["EN-Default", "EN-US", "EN-BR", "EN_INDIA", "EN-AU"]
|
17 |
|
|
|
|
|
18 |
#model = TTS(language='EN', device=device)
|
19 |
def synthesize(language, speaker, text, speed=1.0, progress=gr.Progress()):
|
20 |
model = TTS(language=language, device=device)
|
@@ -23,19 +26,22 @@ def synthesize(language, speaker, text, speed=1.0, progress=gr.Progress()):
|
|
23 |
model.tts_to_file(text, speaker_ids[speaker], bio, speed=speed, pbar=progress.tqdm, format='wav')
|
24 |
return bio.getvalue()
|
25 |
|
26 |
-
def lang(language):
|
27 |
-
if language == "EN":
|
28 |
-
return gr.update(choices=en, value="EN-Default")
|
29 |
-
else:
|
30 |
-
return gr.update(choices=[language], value=language)
|
31 |
with gr.Blocks() as demo:
|
32 |
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.')
|
33 |
with gr.Group():
|
34 |
-
language = gr.Dropdown(languages, interactive=True, value='EN', label='Language')
|
35 |
-
|
|
|
|
|
|
|
36 |
speed = gr.Slider(label='Speed', minimum=0.1, maximum=10.0, value=1.0, interactive=True, step=0.1)
|
37 |
text = gr.Textbox(label="Text to speak", value='The field of text to speech has seen rapid development recently')
|
38 |
-
language.change(fn=lang, inputs=[language], outputs=[speaker])
|
39 |
btn = gr.Button('Synthesize', variant='primary')
|
40 |
aud = gr.Audio(interactive=False)
|
41 |
btn.click(synthesize, inputs=[language, speaker, text, speed], outputs=[aud])
|
|
|
1 |
import gradio as gr
|
2 |
import os, torch, io
|
3 |
+
import sys
|
4 |
#os.system('python -m unidic download')
|
5 |
from melo.api import TTS
|
6 |
speed = 1.0
|
|
|
16 |
languages = ["EN", "ES", "FR", "ZH", "JP", "KR"]
|
17 |
en = ["EN-Default", "EN-US", "EN-BR", "EN_INDIA", "EN-AU"]
|
18 |
|
19 |
+
LANG = sys.argv[1]
|
20 |
+
|
21 |
#model = TTS(language='EN', device=device)
|
22 |
def synthesize(language, speaker, text, speed=1.0, progress=gr.Progress()):
|
23 |
model = TTS(language=language, device=device)
|
|
|
26 |
model.tts_to_file(text, speaker_ids[speaker], bio, speed=speed, pbar=progress.tqdm, format='wav')
|
27 |
return bio.getvalue()
|
28 |
|
29 |
+
#def lang(language):
|
30 |
+
# if language == "EN":
|
31 |
+
# return gr.update(choices=en, value="EN-Default")
|
32 |
+
# else:
|
33 |
+
# return gr.update(choices=[language], value=language)
|
34 |
with gr.Blocks() as demo:
|
35 |
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.')
|
36 |
with gr.Group():
|
37 |
+
# language = gr.Dropdown(languages, interactive=True, value='EN', label='Language')
|
38 |
+
if LANG == "EN":
|
39 |
+
speaker = gr.Dropdown(en, interactive=True, value='EN-Default', label='Speaker')
|
40 |
+
else:
|
41 |
+
speaker = gr.Dropdown([LANG], interactive=True, value='EN-Default', label='Speaker')
|
42 |
speed = gr.Slider(label='Speed', minimum=0.1, maximum=10.0, value=1.0, interactive=True, step=0.1)
|
43 |
text = gr.Textbox(label="Text to speak", value='The field of text to speech has seen rapid development recently')
|
44 |
+
# language.change(fn=lang, inputs=[language], outputs=[speaker])
|
45 |
btn = gr.Button('Synthesize', variant='primary')
|
46 |
aud = gr.Audio(interactive=False)
|
47 |
btn.click(synthesize, inputs=[language, speaker, text, speed], outputs=[aud])
|
melo/text/japanese.py
CHANGED
@@ -11,7 +11,8 @@ punctuation = ["!", "?", "…", ",", ".", "'", "-"]
|
|
11 |
try:
|
12 |
import MeCab
|
13 |
except ImportError as e:
|
14 |
-
|
|
|
15 |
from num2words import num2words
|
16 |
|
17 |
_CONVRULES = [
|
@@ -364,7 +365,11 @@ def hira2kata(text: str) -> str:
|
|
364 |
|
365 |
_SYMBOL_TOKENS = set(list("・、。?!"))
|
366 |
_NO_YOMI_TOKENS = set(list("「」『』―()[][]"))
|
367 |
-
|
|
|
|
|
|
|
|
|
368 |
|
369 |
|
370 |
def text2kata(text: str) -> str:
|
@@ -644,4 +649,4 @@ if __name__ == "__main__":
|
|
644 |
# conv = kakasi.getConverter()
|
645 |
# 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
|
646 |
|
647 |
-
# print(katakana_text) # Output: ニーハオセカイ
|
|
|
11 |
try:
|
12 |
import MeCab
|
13 |
except ImportError as e:
|
14 |
+
print("")
|
15 |
+
# raise ImportError("Japanese requires mecab-python3 and unidic-lite.") from e
|
16 |
from num2words import num2words
|
17 |
|
18 |
_CONVRULES = [
|
|
|
365 |
|
366 |
_SYMBOL_TOKENS = set(list("・、。?!"))
|
367 |
_NO_YOMI_TOKENS = set(list("「」『』―()[][]"))
|
368 |
+
|
369 |
+
try:
|
370 |
+
_TAGGER = MeCab.Tagger()
|
371 |
+
except:
|
372 |
+
print("")
|
373 |
|
374 |
|
375 |
def text2kata(text: str) -> str:
|
|
|
649 |
# conv = kakasi.getConverter()
|
650 |
# 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
|
651 |
|
652 |
+
# print(katakana_text) # Output: ニーハオセカイ
|