Spaces:
Build error
Build error
Update app.py
Browse files
app.py
CHANGED
|
@@ -6,25 +6,47 @@ import gradio as gr
|
|
| 6 |
from TTS.utils.manage import ModelManager
|
| 7 |
from TTS.utils.synthesizer import Synthesizer
|
| 8 |
|
| 9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
MAX_TXT_LEN = 800
|
| 11 |
model_path = os.getcwd() + "/best_model.pth"
|
| 12 |
config_path = os.getcwd() + "/config.json"
|
| 13 |
|
| 14 |
-
synthesizer = Synthesizer(
|
| 15 |
-
model_path, config_path
|
| 16 |
-
)
|
| 17 |
|
| 18 |
|
| 19 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
if len(text) > MAX_TXT_LEN:
|
| 21 |
text = text[:MAX_TXT_LEN]
|
| 22 |
print(f"Input text was cutoff since it went over the {MAX_TXT_LEN} character limit.")
|
| 23 |
print(text)
|
| 24 |
|
| 25 |
|
| 26 |
-
|
| 27 |
# synthesize
|
|
|
|
|
|
|
|
|
|
| 28 |
if synthesizer is None:
|
| 29 |
raise NameError("model not found")
|
| 30 |
wavs = synthesizer.tts(text)
|
|
@@ -37,20 +59,28 @@ def tts(text: str):
|
|
| 37 |
description="""
|
| 38 |
This is a demo of persian text to speech model.
|
| 39 |
|
| 40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
|
| 42 |
-
|
|
|
|
|
|
|
| 43 |
|
| 44 |
|
| 45 |
|
| 46 |
"""
|
| 47 |
article= ""
|
| 48 |
examples=[
|
| 49 |
-
["
|
| 50 |
-
["
|
| 51 |
-
["
|
| 52 |
-
["
|
| 53 |
-
["
|
|
|
|
|
|
|
| 54 |
]
|
| 55 |
iface = gr.Interface(
|
| 56 |
fn=tts,
|
|
@@ -58,11 +88,16 @@ iface = gr.Interface(
|
|
| 58 |
gr.Textbox(
|
| 59 |
label="Text",
|
| 60 |
value="زندگی فقط یک بار است؛ از آن به خوبی استفاده کن",
|
| 61 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
],
|
| 63 |
outputs=gr.Audio(label="Output",type='filepath'),
|
| 64 |
examples=examples,
|
| 65 |
-
title="🗣️Persian
|
| 66 |
description=description,
|
| 67 |
article=article,
|
| 68 |
live=False
|
|
|
|
| 6 |
from TTS.utils.manage import ModelManager
|
| 7 |
from TTS.utils.synthesizer import Synthesizer
|
| 8 |
|
| 9 |
+
MODEL_NAMES=[
|
| 10 |
+
"vits-male",
|
| 11 |
+
"vits-female",
|
| 12 |
+
"glowtts-male",
|
| 13 |
+
"glowtts-female"
|
| 14 |
+
]
|
| 15 |
MAX_TXT_LEN = 800
|
| 16 |
model_path = os.getcwd() + "/best_model.pth"
|
| 17 |
config_path = os.getcwd() + "/config.json"
|
| 18 |
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
|
| 21 |
+
from TTS.utils.download import download_url
|
| 22 |
+
modelInfo=[
|
| 23 |
+
["vits-male","best_model_65633.pth","config-0.json","https://huggingface.co/Kamtera/persian-tts-male-vits/resolve/main/"],
|
| 24 |
+
["vits-female","checkpoint_48000.pth","config-2.json","https://huggingface.co/Kamtera/persian-tts-female-vits/resolve/main/"],
|
| 25 |
+
["glowtts-male","best_model_77797.pth","config-1.json","https://huggingface.co/Kamtera/persian-tts-male-glow_tts/resolve/main/"],
|
| 26 |
+
["glowtts-female","best_model.pth","config.json","https://huggingface.co/Kamtera/persian-tts-female-glow_tts/resolve/main/"]
|
| 27 |
+
]
|
| 28 |
+
|
| 29 |
+
for d in modelInfo:
|
| 30 |
+
directory=d[0]
|
| 31 |
+
if not os.path.exists(directory):
|
| 32 |
+
os.makedirs(directory)
|
| 33 |
+
download_url(
|
| 34 |
+
d[3]+d[1],directory,"best_model.pth"
|
| 35 |
+
)
|
| 36 |
+
download_url(
|
| 37 |
+
d[3]+d[2],directory,"config.json"
|
| 38 |
+
)
|
| 39 |
+
def tts(text: str,model_name: str):
|
| 40 |
if len(text) > MAX_TXT_LEN:
|
| 41 |
text = text[:MAX_TXT_LEN]
|
| 42 |
print(f"Input text was cutoff since it went over the {MAX_TXT_LEN} character limit.")
|
| 43 |
print(text)
|
| 44 |
|
| 45 |
|
|
|
|
| 46 |
# synthesize
|
| 47 |
+
synthesizer = Synthesizer(
|
| 48 |
+
model_name+"/best_model.pth", model_name+"/config.json"
|
| 49 |
+
)
|
| 50 |
if synthesizer is None:
|
| 51 |
raise NameError("model not found")
|
| 52 |
wavs = synthesizer.tts(text)
|
|
|
|
| 59 |
description="""
|
| 60 |
This is a demo of persian text to speech model.
|
| 61 |
|
| 62 |
+
Models can be found here:
|
| 63 |
+
https://huggingface.co/Kamtera/persian-tts-female-vits
|
| 64 |
+
https://huggingface.co/Kamtera/persian-tts-male-vits
|
| 65 |
+
https://huggingface.co/Kamtera/persian-tts-male-glow_tts
|
| 66 |
+
https://huggingface.co/Kamtera/persian-tts-female-glow_tts
|
| 67 |
|
| 68 |
+
Models trained on these datasets :
|
| 69 |
+
https://www.kaggle.com/datasets/magnoliasis/persian-tts-dataset
|
| 70 |
+
https://www.kaggle.com/datasets/magnoliasis/persian-tts-dataset-famale
|
| 71 |
|
| 72 |
|
| 73 |
|
| 74 |
"""
|
| 75 |
article= ""
|
| 76 |
examples=[
|
| 77 |
+
["و خداوند شما را با ارسال روح در جسم زندگانی و حیات بخشید","vits-male"],
|
| 78 |
+
["تاجر تو چه تجارت می کنی ، تو را چه که چه تجارت می کنم؟","vits-female"],
|
| 79 |
+
["شیش سیخ جیگر سیخی شیش هزار","vits-female"],
|
| 80 |
+
["سه شیشه شیر ، سه سیر سرشیر","vits-female"],
|
| 81 |
+
["دزدی دزدید ز بز دزدی بزی ، عجب دزدی که دزدید ز بز دزدی بزی","vits-female"],
|
| 82 |
+
["مثنوی یکی از قالب های شعری است ک هر بیت قافیه ی جداگانه دارد","vits-female"],
|
| 83 |
+
["در گلو ماند خس او سالها، چیست آن خس مهر جاه و مالها","vits-female"],
|
| 84 |
]
|
| 85 |
iface = gr.Interface(
|
| 86 |
fn=tts,
|
|
|
|
| 88 |
gr.Textbox(
|
| 89 |
label="Text",
|
| 90 |
value="زندگی فقط یک بار است؛ از آن به خوبی استفاده کن",
|
| 91 |
+
),
|
| 92 |
+
gr.Radio(
|
| 93 |
+
label="Pick a TTS Model ",
|
| 94 |
+
choices=MODEL_NAMES,
|
| 95 |
+
value="vits-female",
|
| 96 |
+
),
|
| 97 |
],
|
| 98 |
outputs=gr.Audio(label="Output",type='filepath'),
|
| 99 |
examples=examples,
|
| 100 |
+
title="🗣️ Persian tts 🗣️",
|
| 101 |
description=description,
|
| 102 |
article=article,
|
| 103 |
live=False
|