Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Commit
·
b15cc82
1
Parent(s):
468e8b7
Modify structure
Browse files- app.py +174 -0
- public/images/banner.png +3 -0
- public/images/chart.png +3 -0
- public/images/diagram.jpg +3 -0
- public/voices/Cherry.mp3 +3 -0
- public/voices/Chicago.mp3 +3 -0
- public/voices/Janus.mp3 +3 -0
- public/voices/Mantis.mp3 +3 -0
- public/voices/Sponge.mp3 +3 -0
- public/voices/Starchild.mp3 +3 -0
- requirements.txt +2 -0
app.py
ADDED
@@ -0,0 +1,174 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import gradio as gr
|
3 |
+
import modal
|
4 |
+
import traceback
|
5 |
+
|
6 |
+
# --- Configuration ---
|
7 |
+
# This is the name of your Modal stub.
|
8 |
+
MODAL_STUB_NAME = "vibevoice-generator"
|
9 |
+
# This is the name of the remote class and method to call.
|
10 |
+
MODAL_FUNCTION_NAME = "VibeVoiceModel.generate_podcast"
|
11 |
+
|
12 |
+
# These lists are now hardcoded because the data lives on the Modal container.
|
13 |
+
# For a more dynamic app, you could create a small Modal function to fetch these lists.
|
14 |
+
AVAILABLE_MODELS = ["VibeVoice-1.5B", "VibeVoice-7B"]
|
15 |
+
AVAILABLE_VOICES = [
|
16 |
+
"en-Alice_woman_bgm", "en-Alice_woman", "en-Carter_man", "en-Frank_man",
|
17 |
+
"en-Maya_woman", "en-Yasser_man", "in-Samuel_man", "zh-Anchen_man_bgm",
|
18 |
+
"zh-Bowen_man", "zh-Xinran_woman"
|
19 |
+
]
|
20 |
+
DEFAULT_SPEAKERS = ['en-Alice_woman', 'en-Carter_man', 'en-Frank_man', 'en-Maya_woman']
|
21 |
+
|
22 |
+
# --- Modal Connection ---
|
23 |
+
try:
|
24 |
+
# This looks up the remote function on Modal
|
25 |
+
# It will raise an error if the app isn't deployed (`modal deploy modal_runner.py`)
|
26 |
+
remote_generate_function = modal.Function.lookup(MODAL_STUB_NAME, MODAL_FUNCTION_NAME)
|
27 |
+
print("Successfully connected to Modal function.")
|
28 |
+
except modal.exception.NotFoundError:
|
29 |
+
print("ERROR: Modal function not found.")
|
30 |
+
print(f"Please deploy the Modal app first by running: modal deploy modal_runner.py")
|
31 |
+
remote_generate_function = None
|
32 |
+
|
33 |
+
# --- Gradio UI Definition ---
|
34 |
+
theme = gr.themes.Ocean(
|
35 |
+
primary_hue="indigo",
|
36 |
+
secondary_hue="fuchsia",
|
37 |
+
neutral_hue="slate",
|
38 |
+
).set(
|
39 |
+
button_large_radius='*radius_sm'
|
40 |
+
)
|
41 |
+
|
42 |
+
def create_demo_interface():
|
43 |
+
with gr.Blocks(
|
44 |
+
title="VibeVoice - Conference Generator",
|
45 |
+
theme=theme,
|
46 |
+
) as interface:
|
47 |
+
gr.HTML("""
|
48 |
+
<div style="width: 100%; margin-bottom: 20px;">
|
49 |
+
<img src="https://huggingface.co/spaces/ACloudCenter/Conference-Generator-VibeVoice/resolve/main/public/images/banner.png"
|
50 |
+
style="width: 100%; height: auto; border-radius: 15px; box-shadow: 0 10px 40px rgba(0,0,0,0.2);"
|
51 |
+
alt="VibeVoice Banner">
|
52 |
+
</div>
|
53 |
+
""")
|
54 |
+
gr.Markdown("## GPU processing is now offloaded to a Modal.com backend!")
|
55 |
+
|
56 |
+
with gr.Tabs():
|
57 |
+
with gr.Tab("Generate"):
|
58 |
+
gr.Markdown("### Generated Conference")
|
59 |
+
complete_audio_output = gr.Audio(
|
60 |
+
label="Complete Conference (Download)",
|
61 |
+
type="numpy",
|
62 |
+
autoplay=False,
|
63 |
+
show_download_button=True,
|
64 |
+
)
|
65 |
+
|
66 |
+
with gr.Row():
|
67 |
+
with gr.Column(scale=1):
|
68 |
+
gr.Markdown("### Conference Settings")
|
69 |
+
model_dropdown = gr.Dropdown(
|
70 |
+
choices=AVAILABLE_MODELS,
|
71 |
+
value=AVAILABLE_MODELS[0],
|
72 |
+
label="Model",
|
73 |
+
)
|
74 |
+
num_speakers = gr.Slider(
|
75 |
+
minimum=1, maximum=4, value=2, step=1,
|
76 |
+
label="Number of Speakers",
|
77 |
+
)
|
78 |
+
|
79 |
+
gr.Markdown("### Speaker Selection")
|
80 |
+
speaker_selections = []
|
81 |
+
for i in range(4):
|
82 |
+
speaker = gr.Dropdown(
|
83 |
+
choices=AVAILABLE_VOICES,
|
84 |
+
value=DEFAULT_SPEAKERS[i] if i < len(DEFAULT_SPEAKERS) else None,
|
85 |
+
label=f"Speaker {i+1}",
|
86 |
+
visible=(i < 2),
|
87 |
+
)
|
88 |
+
speaker_selections.append(speaker)
|
89 |
+
|
90 |
+
with gr.Accordion("Advanced Settings", open=False):
|
91 |
+
cfg_scale = gr.Slider(
|
92 |
+
minimum=1.0, maximum=2.0, value=1.3, step=0.05,
|
93 |
+
label="CFG Scale (Guidance Strength)",
|
94 |
+
)
|
95 |
+
|
96 |
+
with gr.Column(scale=2):
|
97 |
+
gr.Markdown("### Script Input")
|
98 |
+
script_input = gr.Textbox(
|
99 |
+
label="Conversation Script",
|
100 |
+
placeholder="Enter your conference script here...",
|
101 |
+
lines=12,
|
102 |
+
max_lines=20,
|
103 |
+
)
|
104 |
+
generate_btn = gr.Button(
|
105 |
+
"🚀 Generate Conference (on Modal)", size="lg",
|
106 |
+
variant="primary",
|
107 |
+
)
|
108 |
+
log_output = gr.Textbox(
|
109 |
+
label="Generation Log",
|
110 |
+
lines=8, max_lines=15,
|
111 |
+
interactive=False,
|
112 |
+
)
|
113 |
+
|
114 |
+
def update_speaker_visibility(num_speakers):
|
115 |
+
return [gr.update(visible=(i < num_speakers)) for i in range(4)]
|
116 |
+
|
117 |
+
num_speakers.change(
|
118 |
+
fn=update_speaker_visibility,
|
119 |
+
inputs=[num_speakers],
|
120 |
+
outputs=speaker_selections
|
121 |
+
)
|
122 |
+
|
123 |
+
def generate_podcast_wrapper(model_choice, num_speakers_val, script, *speakers_and_params):
|
124 |
+
if remote_generate_function is None:
|
125 |
+
return None, "ERROR: Modal function not deployed. Please contact the space owner."
|
126 |
+
|
127 |
+
# Show a message that we are calling the remote function
|
128 |
+
yield None, "🔄 Calling remote GPU on Modal.com... this may take a moment to start."
|
129 |
+
|
130 |
+
try:
|
131 |
+
speakers = speakers_and_params[:4]
|
132 |
+
cfg_scale_val = speakers_and_params[4]
|
133 |
+
|
134 |
+
# This is the call to the remote Modal function
|
135 |
+
result, log = remote_generate_function.remote(
|
136 |
+
num_speakers=int(num_speakers_val),
|
137 |
+
script=script,
|
138 |
+
speaker_1=speakers[0],
|
139 |
+
speaker_2=speakers[1],
|
140 |
+
speaker_3=speakers[2],
|
141 |
+
speaker_4=speakers[3],
|
142 |
+
cfg_scale=cfg_scale_val,
|
143 |
+
model_name=model_choice
|
144 |
+
)
|
145 |
+
yield result, log
|
146 |
+
except Exception as e:
|
147 |
+
tb = traceback.format_exc()
|
148 |
+
print(f"Error calling Modal: {e}")
|
149 |
+
yield None, f"An error occurred in the Gradio wrapper: {e}\n\n{tb}"
|
150 |
+
|
151 |
+
generate_btn.click(
|
152 |
+
fn=generate_podcast_wrapper,
|
153 |
+
inputs=[model_dropdown, num_speakers, script_input] + speaker_selections + [cfg_scale],
|
154 |
+
outputs=[complete_audio_output, log_output]
|
155 |
+
)
|
156 |
+
return interface
|
157 |
+
|
158 |
+
# --- Main Execution ---
|
159 |
+
if __name__ == "__main__":
|
160 |
+
if remote_generate_function is None:
|
161 |
+
# If Modal isn't set up, we can't launch the full app.
|
162 |
+
# We'll show a simplified UI with an error message.
|
163 |
+
with gr.Blocks(theme=theme) as interface:
|
164 |
+
gr.Markdown("# ❌ Configuration Error")
|
165 |
+
gr.Markdown(
|
166 |
+
"The Gradio application cannot connect to the Modal backend. "
|
167 |
+
"The Modal app has not been deployed yet. "
|
168 |
+
"Please run `modal deploy modal_runner.py` in your terminal and then refresh this page."
|
169 |
+
)
|
170 |
+
interface.launch()
|
171 |
+
else:
|
172 |
+
# Launch the full Gradio interface
|
173 |
+
interface = create_demo_interface()
|
174 |
+
interface.queue().launch(show_error=True)
|
public/images/banner.png
ADDED
![]() |
Git LFS Details
|
public/images/chart.png
ADDED
![]() |
Git LFS Details
|
public/images/diagram.jpg
ADDED
![]() |
Git LFS Details
|
public/voices/Cherry.mp3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0ac5796a412b2a4e191ecc037be6467cfd3c6233cf7dd2ce07d7742eb046cfbe
|
3 |
+
size 509537
|
public/voices/Chicago.mp3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6be99cc80c226a150402e0ec92ac40206de08ff7399136fc309597b0bd8837ad
|
3 |
+
size 696782
|
public/voices/Janus.mp3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4b8d608ec67b43a088e96691c009a095e488523a6aca29e1fa30d628e46e3ce7
|
3 |
+
size 528763
|
public/voices/Mantis.mp3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e37695776f1c93c4c86c7ba1b26ea365af566363432a02cbd65ca571f99e3c14
|
3 |
+
size 619460
|
public/voices/Sponge.mp3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c7ab366e71ec9c709d49cdb2778bf76f2de27cbbabb1d5452675b2b9a6c24d01
|
3 |
+
size 535450
|
public/voices/Starchild.mp3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e4255d53586fae89e44ba64f13857cc72c823acfa013c47617b42425bfe96d1f
|
3 |
+
size 624894
|
requirements.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
gradio
|
2 |
+
modal
|