Spaces:
Running
Running
Commit
·
c5dd1ce
1
Parent(s):
2f67766
voicebloom
Browse files
app.py
CHANGED
@@ -1,57 +1,94 @@
|
|
1 |
-
import spaces
|
2 |
-
from kokoro import KModel, KPipeline
|
3 |
-
import gradio as gr
|
4 |
import os
|
5 |
import random
|
6 |
import torch
|
|
|
|
|
7 |
|
8 |
-
|
9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
pipelines = {lang_code: KPipeline(lang_code=lang_code, model=False) for lang_code in 'ab'}
|
11 |
pipelines['a'].g2p.lexicon.golds['kokoro'] = 'kˈOkəɹO'
|
12 |
pipelines['b'].g2p.lexicon.golds['kokoro'] = 'kˈQkəɹQ'
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
|
|
|
|
28 |
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
first = True
|
33 |
-
for _, ps, _ in pipeline(text, voice, speed):
|
34 |
-
ref_s = pack[len(ps)-1]
|
35 |
-
audio = models[False](ps, ref_s, speed)
|
36 |
-
yield 24000, audio.numpy()
|
37 |
-
if first:
|
38 |
-
first = False
|
39 |
-
yield 24000, torch.zeros(1).numpy()
|
40 |
|
41 |
-
#
|
42 |
QUOTES = [
|
43 |
"When it is obvious that the goals cannot be reached, don't adjust the goals, adjust the action steps. - Confucius",
|
44 |
"The man who moves a mountain begins by carrying away small stones. - Confucius",
|
45 |
"Life is really simple, but we insist on making it complicated. - Confucius",
|
46 |
"It does not matter how slowly you go as long as you do not stop. - Confucius",
|
|
|
|
|
47 |
"Eating fruit daily provides essential vitamins, minerals and fiber that help maintain good health.",
|
48 |
"An apple a day keeps the doctor away - a simple habit with profound health benefits.",
|
49 |
"Fruits are nature's candy - sweet, nutritious, and vital for our wellbeing.",
|
50 |
"Regular consumption of fruits boosts your immune system and reduces risk of chronic diseases.",
|
51 |
"Colorful fruits on your plate mean a rainbow of nutrients for your body.",
|
52 |
-
"The wisdom of health lies in eating seasonal fruits that nature provides us."
|
|
|
|
|
|
|
53 |
]
|
54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
def get_random_quote():
|
56 |
return random.choice(QUOTES)
|
57 |
|
@@ -77,79 +114,89 @@ Eating seasonal fruits connects us with nature's rhythm and ensures we get the f
|
|
77 |
|
78 |
Remember: 'Nature's first green is gold' - and nowhere is this more true than in the vibrant colors of fruit that nourish our bodies daily."""
|
79 |
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
'🇺🇸 🚺 Kore': 'af_kore',
|
86 |
-
'🇺🇸 🚺 Sarah': 'af_sarah',
|
87 |
-
'🇺🇸 🚹 Michael': 'am_michael',
|
88 |
-
'🇺🇸 🚹 Fenrir': 'am_fenrir',
|
89 |
-
'🇺🇸 🚹 Puck': 'am_puck',
|
90 |
-
'🇺🇸 🚹 Echo': 'am_echo',
|
91 |
-
'🇬🇧 🚺 Emma': 'bf_emma',
|
92 |
-
'🇬🇧 🚺 Isabella': 'bf_isabella',
|
93 |
-
'🇬🇧 🚹 George': 'bm_george',
|
94 |
-
'🇬🇧 🚹 Fable': 'bm_fable',
|
95 |
-
}
|
96 |
-
for v in CHOICES.values():
|
97 |
-
pipelines[v[0]].load_voice(v)
|
98 |
-
|
99 |
-
TOKEN_NOTE = '''
|
100 |
-
💡 Customize pronunciation with Markdown link syntax and /slashes/ like `[Kokoro](/kˈOkəɹO/)`
|
101 |
-
💬 To adjust intonation, try punctuation `;:,.!?—…"()""` or stress `ˈ` and `ˌ`
|
102 |
-
⬇️ Lower stress `[1 level](-1)` or `[2 levels](-2)`
|
103 |
-
⬆️ Raise stress 1 level `[or](+2)` 2 levels (only works on less stressed, usually short words)
|
104 |
-
'''
|
105 |
-
|
106 |
-
with gr.Blocks() as generate_tab:
|
107 |
-
out_audio = gr.Audio(label='Output Audio', interactive=False, streaming=False, autoplay=True)
|
108 |
-
generate_btn = gr.Button('Generate', variant='primary')
|
109 |
-
with gr.Accordion('Output Tokens', open=True):
|
110 |
-
out_ps = gr.Textbox(interactive=False, show_label=False, info='Tokens used to generate the audio, up to 510 context length.')
|
111 |
-
tokenize_btn = gr.Button('Tokenize', variant='secondary')
|
112 |
-
gr.Markdown(TOKEN_NOTE)
|
113 |
-
|
114 |
-
STREAM_NOTE = ['⚠️ There is an unknown Gradio bug that might yield no audio the first time you click `Stream`.']
|
115 |
-
STREAM_NOTE = '\n\n'.join(STREAM_NOTE)
|
116 |
-
|
117 |
-
with gr.Blocks() as stream_tab:
|
118 |
-
out_stream = gr.Audio(label='Output Audio Stream', interactive=False, streaming=True, autoplay=True)
|
119 |
-
with gr.Row():
|
120 |
-
stream_btn = gr.Button('Stream', variant='primary')
|
121 |
-
stop_btn = gr.Button('Stop', variant='stop')
|
122 |
-
with gr.Accordion('Note', open=True):
|
123 |
-
gr.Markdown(STREAM_NOTE)
|
124 |
-
|
125 |
-
BANNER_TEXT = '''
|
126 |
-
# Kokoro TTS - Wisdom Speaker
|
127 |
-
|
128 |
-
This demo uses the open-weight Kokoro TTS model to convert wisdom quotes into speech.
|
129 |
-
'''
|
130 |
-
|
131 |
-
with gr.Blocks() as app:
|
132 |
-
with gr.Row():
|
133 |
-
gr.Markdown(BANNER_TEXT, container=True)
|
134 |
with gr.Row():
|
135 |
with gr.Column():
|
136 |
-
|
137 |
-
with gr.
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
144 |
with gr.Column():
|
145 |
-
gr.
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
152 |
stop_btn.click(fn=None, cancels=stream_event)
|
153 |
|
154 |
-
if __name__ ==
|
155 |
-
app.
|
|
|
|
|
|
|
|
|
1 |
import os
|
2 |
import random
|
3 |
import torch
|
4 |
+
from kokoro import KModel, KPipeline
|
5 |
+
import gradio as gr
|
6 |
|
7 |
+
# App configuration
|
8 |
+
APP_TITLE = "✨ VoiceBloom ✨"
|
9 |
+
APP_SUBTITLE = "Transform wisdom into delightful speech!"
|
10 |
+
THEME = gr.themes.Soft(
|
11 |
+
primary_hue="indigo",
|
12 |
+
secondary_hue="purple",
|
13 |
+
).set(
|
14 |
+
body_background_fill="linear-gradient(to right top, #d16ba5, #c777b9, #ba83ca, #aa8fd8, #9a9ae1, #8aa7ec, #79b3f4, #69bff8, #52cffe, #41dfff, #46eefa, #5ffbf1)",
|
15 |
+
button_primary_background_fill="linear-gradient(90deg, rgba(255,124,0,1) 0%, rgba(255,194,23,1) 100%)",
|
16 |
+
button_primary_background_fill_hover="linear-gradient(90deg, rgba(255,194,23,1) 0%, rgba(255,124,0,1) 100%)",
|
17 |
+
button_secondary_background_fill="linear-gradient(90deg, rgba(144,95,255,1) 0%, rgba(110,72,220,1) 100%)",
|
18 |
+
button_secondary_background_fill_hover="linear-gradient(90deg, rgba(110,72,220,1) 0%, rgba(144,95,255,1) 100%)",
|
19 |
+
block_background_fill="rgba(255, 255, 255, 0.8)",
|
20 |
+
block_shadow="0px 4px 12px rgba(0, 0, 0, 0.1)",
|
21 |
+
block_radius="12px",
|
22 |
+
)
|
23 |
+
|
24 |
+
# Initialize TTS models without GPU
|
25 |
+
models = {False: KModel().to('cpu').eval()}
|
26 |
pipelines = {lang_code: KPipeline(lang_code=lang_code, model=False) for lang_code in 'ab'}
|
27 |
pipelines['a'].g2p.lexicon.golds['kokoro'] = 'kˈOkəɹO'
|
28 |
pipelines['b'].g2p.lexicon.golds['kokoro'] = 'kˈQkəɹQ'
|
29 |
|
30 |
+
# Define voice options with fun emoji indicators
|
31 |
+
VOICES = {
|
32 |
+
'✨ Sarah (US) ✨': 'af_sarah',
|
33 |
+
'🌟 Nicole (US) 🌟': 'af_nicole',
|
34 |
+
'💖 Heart (US) 💖': 'af_heart',
|
35 |
+
'🔮 Bella (US) 🔮': 'af_bella',
|
36 |
+
'🌈 Aoede (US) 🌈': 'af_aoede',
|
37 |
+
'🎵 Michael (US) 🎵': 'am_michael',
|
38 |
+
'🌠 Echo (US) 🌠': 'am_echo',
|
39 |
+
'🧙 Fenrir (US) 🧙': 'am_fenrir',
|
40 |
+
'🎭 Puck (US) 🎭': 'am_puck',
|
41 |
+
'👑 Emma (UK) 👑': 'bf_emma',
|
42 |
+
'🌹 Isabella (UK) 🌹': 'bf_isabella',
|
43 |
+
'🎩 George (UK) 🎩': 'bm_george',
|
44 |
+
'✨ Fable (UK) ✨': 'bm_fable',
|
45 |
+
}
|
46 |
|
47 |
+
# Preload voices
|
48 |
+
for v in VOICES.values():
|
49 |
+
pipelines[v[0]].load_voice(v)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
|
51 |
+
# Inspirational quotes
|
52 |
QUOTES = [
|
53 |
"When it is obvious that the goals cannot be reached, don't adjust the goals, adjust the action steps. - Confucius",
|
54 |
"The man who moves a mountain begins by carrying away small stones. - Confucius",
|
55 |
"Life is really simple, but we insist on making it complicated. - Confucius",
|
56 |
"It does not matter how slowly you go as long as you do not stop. - Confucius",
|
57 |
+
"Our greatest glory is not in never falling, but in rising every time we fall. - Confucius",
|
58 |
+
"Silence is a true friend who never betrays. - Confucius",
|
59 |
"Eating fruit daily provides essential vitamins, minerals and fiber that help maintain good health.",
|
60 |
"An apple a day keeps the doctor away - a simple habit with profound health benefits.",
|
61 |
"Fruits are nature's candy - sweet, nutritious, and vital for our wellbeing.",
|
62 |
"Regular consumption of fruits boosts your immune system and reduces risk of chronic diseases.",
|
63 |
"Colorful fruits on your plate mean a rainbow of nutrients for your body.",
|
64 |
+
"The wisdom of health lies in eating seasonal fruits that nature provides us.",
|
65 |
+
"A journey of a thousand miles begins with a single step.",
|
66 |
+
"Happiness is not something ready-made. It comes from your own actions.",
|
67 |
+
"The best time to plant a tree was 20 years ago. The second best time is now.",
|
68 |
]
|
69 |
|
70 |
+
def generate_audio(text, voice, speed):
|
71 |
+
"""Generate audio from text using the selected voice and speed"""
|
72 |
+
pipeline = pipelines[voice[0]]
|
73 |
+
pack = pipeline.load_voice(voice)
|
74 |
+
|
75 |
+
for _, ps, _ in pipeline(text, voice, speed):
|
76 |
+
ref_s = pack[len(ps)-1]
|
77 |
+
audio = models[False](ps, ref_s, speed)
|
78 |
+
return (24000, audio.numpy())
|
79 |
+
|
80 |
+
return None
|
81 |
+
|
82 |
+
def stream_audio(text, voice, speed):
|
83 |
+
"""Stream audio for longer texts"""
|
84 |
+
pipeline = pipelines[voice[0]]
|
85 |
+
pack = pipeline.load_voice(voice)
|
86 |
+
|
87 |
+
for _, ps, _ in pipeline(text, voice, speed):
|
88 |
+
ref_s = pack[len(ps)-1]
|
89 |
+
audio = models[False](ps, ref_s, speed)
|
90 |
+
yield 24000, audio.numpy()
|
91 |
+
|
92 |
def get_random_quote():
|
93 |
return random.choice(QUOTES)
|
94 |
|
|
|
114 |
|
115 |
Remember: 'Nature's first green is gold' - and nowhere is this more true than in the vibrant colors of fruit that nourish our bodies daily."""
|
116 |
|
117 |
+
# Build the UI
|
118 |
+
with gr.Blocks(theme=THEME) as app:
|
119 |
+
gr.Markdown(f"# {APP_TITLE}")
|
120 |
+
gr.Markdown(f"### {APP_SUBTITLE}")
|
121 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
122 |
with gr.Row():
|
123 |
with gr.Column():
|
124 |
+
# Input section with colorful styling
|
125 |
+
with gr.Box():
|
126 |
+
gr.Markdown("### 📝 What would you like me to say?")
|
127 |
+
text_input = gr.Textbox(
|
128 |
+
label="",
|
129 |
+
placeholder="Type your text here or choose a wisdom quote below...",
|
130 |
+
lines=5
|
131 |
+
)
|
132 |
+
|
133 |
+
with gr.Row():
|
134 |
+
random_btn = gr.Button("🎲 Random Quote", variant="secondary")
|
135 |
+
confucius_btn = gr.Button("🧙 Confucius Wisdom", variant="secondary")
|
136 |
+
fruit_btn = gr.Button("🍎 Fruit Wisdom", variant="secondary")
|
137 |
+
|
138 |
+
# Voice & Speed controls
|
139 |
+
with gr.Box():
|
140 |
+
gr.Markdown("### 🎤 Choose Your Voice")
|
141 |
+
voice_dropdown = gr.Dropdown(
|
142 |
+
list(VOICES.items()),
|
143 |
+
value=list(VOICES.items())[0][1],
|
144 |
+
label="Voice Style"
|
145 |
+
)
|
146 |
+
|
147 |
+
gr.Markdown("### ⏱️ Adjust Speaking Speed")
|
148 |
+
speed_slider = gr.Slider(
|
149 |
+
minimum=0.5,
|
150 |
+
maximum=2.0,
|
151 |
+
value=1.0,
|
152 |
+
step=0.1,
|
153 |
+
label="Speed"
|
154 |
+
)
|
155 |
+
|
156 |
+
# Output section
|
157 |
with gr.Column():
|
158 |
+
with gr.Tabs():
|
159 |
+
with gr.TabItem("🔊 Listen"):
|
160 |
+
audio_output = gr.Audio(
|
161 |
+
label="Your VoiceBloom Creation",
|
162 |
+
interactive=False,
|
163 |
+
autoplay=True
|
164 |
+
)
|
165 |
+
generate_btn = gr.Button("🎵 Generate Audio", variant="primary", size="lg")
|
166 |
+
|
167 |
+
with gr.TabItem("📻 Stream Longer Text"):
|
168 |
+
stream_output = gr.Audio(
|
169 |
+
label="Streaming Audio",
|
170 |
+
interactive=False,
|
171 |
+
streaming=True,
|
172 |
+
autoplay=True
|
173 |
+
)
|
174 |
+
with gr.Row():
|
175 |
+
stream_btn = gr.Button("▶️ Start Streaming", variant="primary")
|
176 |
+
stop_btn = gr.Button("⏹️ Stop", variant="stop")
|
177 |
+
|
178 |
+
# Footer
|
179 |
+
gr.Markdown("---")
|
180 |
+
gr.Markdown("### ✨ VoiceBloom - Bringing wisdom to life through the art of voice ✨")
|
181 |
+
|
182 |
+
# Set up event handlers
|
183 |
+
random_btn.click(fn=get_random_quote, outputs=[text_input])
|
184 |
+
confucius_btn.click(fn=get_confucius, outputs=[text_input])
|
185 |
+
fruit_btn.click(fn=get_fruit_wisdom, outputs=[text_input])
|
186 |
+
|
187 |
+
generate_btn.click(
|
188 |
+
fn=generate_audio,
|
189 |
+
inputs=[text_input, voice_dropdown, speed_slider],
|
190 |
+
outputs=[audio_output]
|
191 |
+
)
|
192 |
+
|
193 |
+
stream_event = stream_btn.click(
|
194 |
+
fn=stream_audio,
|
195 |
+
inputs=[text_input, voice_dropdown, speed_slider],
|
196 |
+
outputs=[stream_output]
|
197 |
+
)
|
198 |
+
|
199 |
stop_btn.click(fn=None, cancels=stream_event)
|
200 |
|
201 |
+
if __name__ == "__main__":
|
202 |
+
app.launch()
|