ui updated
Browse files
app.py
CHANGED
@@ -59,132 +59,175 @@ def process_audio(audio_path, model_name):
|
|
59 |
real_prob = 1 - prob
|
60 |
fake_prob = prob
|
61 |
|
62 |
-
# Return formatted results
|
63 |
return {
|
64 |
-
"
|
65 |
-
"
|
66 |
}
|
67 |
|
68 |
except Exception as e:
|
69 |
-
return {"
|
70 |
|
71 |
|
72 |
def predict(audio_file, model_name):
|
73 |
"""Gradio interface function"""
|
74 |
if audio_file is None:
|
75 |
-
return {"
|
76 |
return process_audio(audio_file, model_name)
|
77 |
|
78 |
|
79 |
-
# Custom CSS for styling
|
80 |
css = """
|
81 |
:root {
|
82 |
-
--primary-
|
83 |
-
--secondary-
|
84 |
-
--
|
85 |
-
--
|
86 |
-
--text-
|
87 |
-
--border-
|
|
|
|
|
|
|
88 |
}
|
89 |
|
90 |
-
.gradio-container {
|
91 |
-
background-color: var(--
|
|
|
92 |
}
|
93 |
|
94 |
-
.
|
95 |
-
background
|
96 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
97 |
color: white !important;
|
98 |
-
border
|
|
|
|
|
|
|
|
|
99 |
}
|
100 |
|
101 |
-
|
102 |
-
background:
|
103 |
-
|
104 |
-
|
105 |
-
transition: all 0.
|
106 |
}
|
107 |
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
|
|
|
|
112 |
}
|
113 |
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
|
|
|
|
119 |
}
|
120 |
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
127 |
}
|
128 |
|
129 |
-
.
|
130 |
-
|
131 |
-
|
132 |
-
|
|
|
133 |
}
|
134 |
|
135 |
.header-container {
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
border
|
141 |
-
box-shadow: 0 4px 15px rgba(0, 0, 0, 0.05);
|
142 |
}
|
143 |
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
}
|
151 |
|
152 |
-
.
|
153 |
-
|
154 |
-
padding:
|
155 |
-
|
156 |
-
|
157 |
-
color: var(--primary-color);
|
158 |
-
text-decoration: none;
|
159 |
-
box-shadow: 0 2px 5px rgba(0, 0, 0, 0.1);
|
160 |
-
transition: all 0.2s ease;
|
161 |
}
|
162 |
|
163 |
-
|
164 |
-
|
165 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
166 |
}
|
167 |
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
172 |
}
|
173 |
"""
|
174 |
|
175 |
# Create Gradio interface
|
176 |
-
with gr.Blocks(css=css) as demo:
|
177 |
-
# Title
|
178 |
gr.HTML(
|
179 |
"""
|
180 |
<div class="header-container">
|
181 |
-
<div
|
182 |
<img src="https://i.postimg.cc/3Jx3yZ5b/real-vs-fake-sonics-w-logo.jpg"
|
183 |
-
style="max-width:
|
184 |
</div>
|
185 |
-
<h1
|
186 |
-
<h3>ICLR 2025 [Poster]</h3>
|
187 |
-
<p style="
|
188 |
Detect if a song is real or AI-generated with our state-of-the-art models.
|
189 |
Simply upload an audio file to verify its authenticity!
|
190 |
</p>
|
@@ -192,10 +235,10 @@ with gr.Blocks(css=css) as demo:
|
|
192 |
"""
|
193 |
)
|
194 |
|
195 |
-
# Resource Links
|
196 |
gr.HTML(
|
197 |
"""
|
198 |
-
<div
|
199 |
<a href="https://openreview.net/forum?id=PY7KSh29Z8" target="_blank" class="resource-link">
|
200 |
📄 Paper
|
201 |
</a>
|
@@ -219,15 +262,16 @@ with gr.Blocks(css=css) as demo:
|
|
219 |
with gr.Row(equal_height=True):
|
220 |
with gr.Column():
|
221 |
audio_input = gr.Audio(
|
222 |
-
label="
|
223 |
type="filepath",
|
224 |
-
elem_id="audio_input"
|
|
|
225 |
)
|
226 |
|
227 |
model_dropdown = gr.Dropdown(
|
228 |
choices=list(MODEL_IDS.keys()),
|
229 |
value="SpecTTTra-γ (5s)",
|
230 |
-
label="
|
231 |
elem_id="model_dropdown"
|
232 |
)
|
233 |
|
@@ -239,33 +283,33 @@ with gr.Blocks(css=css) as demo:
|
|
239 |
with gr.Column():
|
240 |
# Define output before using it in Examples
|
241 |
output = gr.Label(
|
242 |
-
label="
|
243 |
num_top_classes=2,
|
244 |
-
elem_id="output"
|
245 |
-
elem_classes="label-container"
|
246 |
)
|
247 |
|
248 |
-
with gr.Accordion("
|
249 |
gr.Markdown("""
|
250 |
-
The SONICS classifier
|
251 |
-
|
|
|
252 |
|
253 |
-
|
254 |
-
- **SpecTTTra
|
255 |
-
- **SpecTTTra-β**: Balanced performance
|
256 |
-
- **SpecTTTra
|
257 |
|
258 |
-
|
259 |
- **5s**: Analyzes a 5-second clip (faster)
|
260 |
- **120s**: Analyzes up to 2 minutes (more accurate)
|
261 |
""")
|
262 |
|
263 |
# Add Examples section after output is defined
|
264 |
-
with gr.Accordion("
|
265 |
gr.Examples(
|
266 |
examples=[
|
267 |
-
["
|
268 |
-
["
|
269 |
],
|
270 |
inputs=[audio_input, model_dropdown],
|
271 |
outputs=[output],
|
@@ -276,9 +320,9 @@ with gr.Blocks(css=css) as demo:
|
|
276 |
# Footer
|
277 |
gr.HTML(
|
278 |
"""
|
279 |
-
<div class="footer">
|
280 |
-
<p>SONICS: Synthetic Or Not - Identifying Counterfeit Songs |
|
281 |
-
<p
|
282 |
</div>
|
283 |
"""
|
284 |
)
|
@@ -287,4 +331,5 @@ with gr.Blocks(css=css) as demo:
|
|
287 |
submit_btn.click(fn=predict, inputs=[audio_input, model_dropdown], outputs=[output])
|
288 |
|
289 |
if __name__ == "__main__":
|
|
|
290 |
demo.launch()
|
|
|
59 |
real_prob = 1 - prob
|
60 |
fake_prob = prob
|
61 |
|
62 |
+
# Return formatted results
|
63 |
return {
|
64 |
+
"Real": float(real_prob),
|
65 |
+
"Fake": float(fake_prob)
|
66 |
}
|
67 |
|
68 |
except Exception as e:
|
69 |
+
return {"Error": str(e)}
|
70 |
|
71 |
|
72 |
def predict(audio_file, model_name):
|
73 |
"""Gradio interface function"""
|
74 |
if audio_file is None:
|
75 |
+
return {"Message": "Please upload an audio file"}
|
76 |
return process_audio(audio_file, model_name)
|
77 |
|
78 |
|
79 |
+
# Custom CSS for styling - Dark theme with black background
|
80 |
css = """
|
81 |
:root {
|
82 |
+
--primary-bg: #000000;
|
83 |
+
--secondary-bg: #111111;
|
84 |
+
--panel-bg: #1e1e1e;
|
85 |
+
--text-color: #ffffff;
|
86 |
+
--text-secondary: #bbbbbb;
|
87 |
+
--border-color: #333333;
|
88 |
+
--analyze-button-color: #ffa500;
|
89 |
+
--analyze-button-hover: #ff8c00;
|
90 |
+
--accent-color: #4a78e5;
|
91 |
}
|
92 |
|
93 |
+
body, .gradio-container {
|
94 |
+
background-color: var(--primary-bg) !important;
|
95 |
+
color: var(--text-color) !important;
|
96 |
}
|
97 |
|
98 |
+
.footer, .header-container, .accordion-content {
|
99 |
+
background-color: var(--secondary-bg) !important;
|
100 |
+
color: var(--text-color) !important;
|
101 |
+
}
|
102 |
+
|
103 |
+
/* Headers and text */
|
104 |
+
h1, h2, h3 {
|
105 |
+
color: var(--text-color) !important;
|
106 |
+
}
|
107 |
+
|
108 |
+
p {
|
109 |
+
color: var(--text-secondary) !important;
|
110 |
+
}
|
111 |
+
|
112 |
+
/* Button styling */
|
113 |
+
button#submit_btn {
|
114 |
+
background-color: var(--analyze-button-color) !important;
|
115 |
color: white !important;
|
116 |
+
border: none !important;
|
117 |
+
font-weight: bold !important;
|
118 |
+
padding: 10px 20px !important;
|
119 |
+
font-size: 16px !important;
|
120 |
+
border-radius: 8px !important;
|
121 |
}
|
122 |
|
123 |
+
button#submit_btn:hover {
|
124 |
+
background-color: var(--analyze-button-hover) !important;
|
125 |
+
box-shadow: 0 4px 8px rgba(0, 0, 0, 0.5) !important;
|
126 |
+
transform: translateY(-2px) !important;
|
127 |
+
transition: all 0.2s ease !important;
|
128 |
}
|
129 |
|
130 |
+
/* Panel backgrounds */
|
131 |
+
.gr-panel, .gr-box, .gr-form, .gr-input-label, .gr-input {
|
132 |
+
background-color: var(--panel-bg) !important;
|
133 |
+
border: 1px solid var(--border-color) !important;
|
134 |
+
border-radius: 8px !important;
|
135 |
+
color: var(--text-color) !important;
|
136 |
}
|
137 |
|
138 |
+
/* Results panel */
|
139 |
+
#output {
|
140 |
+
background-color: var(--panel-bg) !important;
|
141 |
+
border-radius: 8px !important;
|
142 |
+
padding: 10px !important;
|
143 |
+
box-shadow: 0 2px 5px rgba(0, 0, 0, 0.3) !important;
|
144 |
+
border: 1px solid var(--border-color) !important;
|
145 |
}
|
146 |
|
147 |
+
/* Resource links */
|
148 |
+
.resource-link {
|
149 |
+
background-color: var(--secondary-bg) !important;
|
150 |
+
color: var(--accent-color) !important;
|
151 |
+
border: 1px solid var(--border-color) !important;
|
152 |
+
padding: 8px 16px !important;
|
153 |
+
border-radius: 20px !important;
|
154 |
+
margin: 5px !important;
|
155 |
+
text-decoration: none !important;
|
156 |
+
display: inline-block !important;
|
157 |
+
font-weight: 500 !important;
|
158 |
+
box-shadow: 0 2px 4px rgba(0, 0, 0, 0.3) !important;
|
159 |
}
|
160 |
|
161 |
+
.resource-link:hover {
|
162 |
+
transform: translateY(-2px) !important;
|
163 |
+
box-shadow: 0 3px 6px rgba(0, 0, 0, 0.4) !important;
|
164 |
+
transition: all 0.2s ease !important;
|
165 |
+
background-color: #222222 !important;
|
166 |
}
|
167 |
|
168 |
.header-container {
|
169 |
+
padding: 20px !important;
|
170 |
+
border-radius: 10px !important;
|
171 |
+
box-shadow: 0 2px 10px rgba(0, 0, 0, 0.3) !important;
|
172 |
+
margin-bottom: 20px !important;
|
173 |
+
border: 1px solid var(--border-color) !important;
|
|
|
174 |
}
|
175 |
|
176 |
+
/* Accordion styling */
|
177 |
+
.gr-accordion {
|
178 |
+
border: 1px solid var(--border-color) !important;
|
179 |
+
border-radius: 8px !important;
|
180 |
+
overflow: hidden !important;
|
181 |
+
background-color: var(--panel-bg) !important;
|
182 |
}
|
183 |
|
184 |
+
.gr-accordion-header {
|
185 |
+
background-color: var(--secondary-bg) !important;
|
186 |
+
padding: 10px 15px !important;
|
187 |
+
font-weight: 600 !important;
|
188 |
+
color: var(--text-color) !important;
|
|
|
|
|
|
|
|
|
189 |
}
|
190 |
|
191 |
+
/* Audio player */
|
192 |
+
.audio-player {
|
193 |
+
background-color: var(--panel-bg) !important;
|
194 |
+
border-radius: 8px !important;
|
195 |
+
overflow: hidden !important;
|
196 |
+
}
|
197 |
+
|
198 |
+
/* Dropdown & Input fields */
|
199 |
+
select, input, .gr-dropdown {
|
200 |
+
background-color: var(--panel-bg) !important;
|
201 |
+
color: var(--text-color) !important;
|
202 |
+
border: 1px solid var(--border-color) !important;
|
203 |
}
|
204 |
|
205 |
+
/* Labels */
|
206 |
+
label, .gr-label {
|
207 |
+
color: var(--text-secondary) !important;
|
208 |
+
}
|
209 |
+
|
210 |
+
/* Footer styling */
|
211 |
+
.footer {
|
212 |
+
border-top: 1px solid var(--border-color) !important;
|
213 |
+
margin-top: 30px !important;
|
214 |
+
padding: 15px !important;
|
215 |
}
|
216 |
"""
|
217 |
|
218 |
# Create Gradio interface
|
219 |
+
with gr.Blocks(css=css, theme=gr.themes.Default()) as demo:
|
220 |
+
# Title and Logo
|
221 |
gr.HTML(
|
222 |
"""
|
223 |
<div class="header-container">
|
224 |
+
<div style="display: flex; justify-content: center; margin-bottom: 20px;">
|
225 |
<img src="https://i.postimg.cc/3Jx3yZ5b/real-vs-fake-sonics-w-logo.jpg"
|
226 |
+
style="max-width: 150px; border-radius: 10px; box-shadow: 0 4px 8px rgba(0,0,0,0.3);">
|
227 |
</div>
|
228 |
+
<h1 style="text-align: center; font-size: 28px; margin-bottom: 10px; color: #ffffff;">SONICS: Synthetic Or Not - Identifying Counterfeit Songs</h1>
|
229 |
+
<h3 style="text-align: center; color: #bbbbbb; margin-bottom: 15px;">ICLR 2025 [Poster]</h3>
|
230 |
+
<p style="text-align: center; font-size: 16px; margin: 0; color: #aaaaaa;">
|
231 |
Detect if a song is real or AI-generated with our state-of-the-art models.
|
232 |
Simply upload an audio file to verify its authenticity!
|
233 |
</p>
|
|
|
235 |
"""
|
236 |
)
|
237 |
|
238 |
+
# Resource Links - Simplified
|
239 |
gr.HTML(
|
240 |
"""
|
241 |
+
<div style="display: flex; justify-content: center; flex-wrap: wrap; gap: 8px; margin-bottom: 25px;">
|
242 |
<a href="https://openreview.net/forum?id=PY7KSh29Z8" target="_blank" class="resource-link">
|
243 |
📄 Paper
|
244 |
</a>
|
|
|
262 |
with gr.Row(equal_height=True):
|
263 |
with gr.Column():
|
264 |
audio_input = gr.Audio(
|
265 |
+
label="Upload Audio File",
|
266 |
type="filepath",
|
267 |
+
elem_id="audio_input",
|
268 |
+
elem_classes="audio-player"
|
269 |
)
|
270 |
|
271 |
model_dropdown = gr.Dropdown(
|
272 |
choices=list(MODEL_IDS.keys()),
|
273 |
value="SpecTTTra-γ (5s)",
|
274 |
+
label="Select Model",
|
275 |
elem_id="model_dropdown"
|
276 |
)
|
277 |
|
|
|
283 |
with gr.Column():
|
284 |
# Define output before using it in Examples
|
285 |
output = gr.Label(
|
286 |
+
label="Analysis Result",
|
287 |
num_top_classes=2,
|
288 |
+
elem_id="output"
|
|
|
289 |
)
|
290 |
|
291 |
+
with gr.Accordion("How It Works", open=True):
|
292 |
gr.Markdown("""
|
293 |
+
## The SONICS classifier
|
294 |
+
|
295 |
+
The SONICS classifier analyzes your audio to determine if it's an authentic song (human created) or generated by AI. Our models are trained on a diverse dataset of real and AI-generated songs from Suno and Udio.
|
296 |
|
297 |
+
### Models available:
|
298 |
+
- **SpecTTTra-α**: Optimized for speed
|
299 |
+
- **SpecTTTra-β**: Balanced performance
|
300 |
+
- **SpecTTTra-γ**: Highest accuracy
|
301 |
|
302 |
+
### Duration variants:
|
303 |
- **5s**: Analyzes a 5-second clip (faster)
|
304 |
- **120s**: Analyzes up to 2 minutes (more accurate)
|
305 |
""")
|
306 |
|
307 |
# Add Examples section after output is defined
|
308 |
+
with gr.Accordion("Example Audio Files", open=True):
|
309 |
gr.Examples(
|
310 |
examples=[
|
311 |
+
["demo/real_song.mp3", "SpecTTTra-γ (5s)"],
|
312 |
+
["demo/fake_song.mp3", "SpecTTTra-γ (5s)"],
|
313 |
],
|
314 |
inputs=[audio_input, model_dropdown],
|
315 |
outputs=[output],
|
|
|
320 |
# Footer
|
321 |
gr.HTML(
|
322 |
"""
|
323 |
+
<div class="footer" style="text-align: center;">
|
324 |
+
<p style="color: #bbbbbb; font-size: 14px;">SONICS: Synthetic Or Not - Identifying Counterfeit Songs | ICLR 2025</p>
|
325 |
+
<p style="color: #777777; font-size: 12px;">For research purposes only</p>
|
326 |
</div>
|
327 |
"""
|
328 |
)
|
|
|
331 |
submit_btn.click(fn=predict, inputs=[audio_input, model_dropdown], outputs=[output])
|
332 |
|
333 |
if __name__ == "__main__":
|
334 |
+
# Use dark theme as base and then apply custom CSS on top
|
335 |
demo.launch()
|