Spaces:
Sleeping
Sleeping
Commit
·
deb14ad
1
Parent(s):
97e4faf
Update app.py
Browse files
app.py
CHANGED
@@ -15,7 +15,6 @@ VOICE_OPTIONS = [
|
|
15 |
"random", # special option for random voice
|
16 |
]
|
17 |
|
18 |
-
|
19 |
def inference(
|
20 |
text,
|
21 |
script,
|
@@ -46,7 +45,6 @@ def inference(
|
|
46 |
|
47 |
start_time = time.time()
|
48 |
|
49 |
-
# all_parts = []
|
50 |
for j, text in enumerate(texts):
|
51 |
for audio_frame in tts.tts_with_preset(
|
52 |
text,
|
@@ -55,26 +53,11 @@ def inference(
|
|
55 |
preset="ultra_fast",
|
56 |
k=1
|
57 |
):
|
58 |
-
# print("Time taken: ", time.time() - start_time)
|
59 |
-
# all_parts.append(audio_frame)
|
60 |
yield (24000, audio_frame.cpu().detach().numpy())
|
61 |
|
62 |
-
# wav = torch.cat(all_parts, dim=0).unsqueeze(0)
|
63 |
-
# print(wav.shape)
|
64 |
-
# torchaudio.save("output.wav", wav.cpu(), 24000)
|
65 |
-
# yield (None, gr.make_waveform(audio="output.wav",))
|
66 |
def main():
|
67 |
-
title = "Tortoise TTS
|
68 |
-
|
69 |
-
A text-to-speech system which powers lot of organizations in Speech synthesis domain.
|
70 |
-
<br/>
|
71 |
-
a model with strong multi-voice capabilities, highly realistic prosody and intonation.
|
72 |
-
<br/>
|
73 |
-
for faster inference, use the 'ultra_fast' preset and duplicate space if you don't want to wait in a queue.
|
74 |
-
<br/>
|
75 |
-
"""
|
76 |
-
text = gr.Textbox(
|
77 |
-
lines=4,
|
78 |
label="Text (Provide either text, or upload a newline separated text file below):",
|
79 |
)
|
80 |
script = gr.File(label="Upload a text file")
|
@@ -96,7 +79,6 @@ def main():
|
|
96 |
)
|
97 |
|
98 |
output_audio = gr.Audio(label="streaming audio:", streaming=True, autoplay=True)
|
99 |
-
# download_audio = gr.Audio(label="dowanload audio:")
|
100 |
interface = gr.Interface(
|
101 |
fn=inference,
|
102 |
inputs=[
|
@@ -107,18 +89,6 @@ def main():
|
|
107 |
split_by_newline,
|
108 |
],
|
109 |
title=title,
|
110 |
-
description=description,
|
111 |
outputs=[output_audio],
|
112 |
)
|
113 |
-
interface.queue().launch()
|
114 |
-
|
115 |
-
|
116 |
-
if __name__ == "__main__":
|
117 |
-
tts = TextToSpeech(kv_cache=True, use_deepspeed=True, half=True)
|
118 |
-
|
119 |
-
with open("Tortoise_TTS_Runs_Scripts.log", "a") as f:
|
120 |
-
f.write(
|
121 |
-
f"\n\n-------------------------Tortoise TTS Scripts Logs, {datetime.now()}-------------------------\n"
|
122 |
-
)
|
123 |
-
|
124 |
-
main()
|
|
|
15 |
"random", # special option for random voice
|
16 |
]
|
17 |
|
|
|
18 |
def inference(
|
19 |
text,
|
20 |
script,
|
|
|
45 |
|
46 |
start_time = time.time()
|
47 |
|
|
|
48 |
for j, text in enumerate(texts):
|
49 |
for audio_frame in tts.tts_with_preset(
|
50 |
text,
|
|
|
53 |
preset="ultra_fast",
|
54 |
k=1
|
55 |
):
|
|
|
|
|
56 |
yield (24000, audio_frame.cpu().detach().numpy())
|
57 |
|
|
|
|
|
|
|
|
|
58 |
def main():
|
59 |
+
title = "Tortoise TTS "
|
60 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
label="Text (Provide either text, or upload a newline separated text file below):",
|
62 |
)
|
63 |
script = gr.File(label="Upload a text file")
|
|
|
79 |
)
|
80 |
|
81 |
output_audio = gr.Audio(label="streaming audio:", streaming=True, autoplay=True)
|
|
|
82 |
interface = gr.Interface(
|
83 |
fn=inference,
|
84 |
inputs=[
|
|
|
89 |
split_by_newline,
|
90 |
],
|
91 |
title=title,
|
|
|
92 |
outputs=[output_audio],
|
93 |
)
|
94 |
+
interface.queue().launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|