Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -157,290 +157,82 @@ def process_video(
|
|
| 157 |
return video_summary, frame_previews
|
| 158 |
|
| 159 |
# Create the Gradio interface
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
:
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
|
| 198 |
-
|
| 199 |
-
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
| 203 |
-
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
|
| 207 |
-
|
| 208 |
-
|
| 209 |
-
|
| 210 |
-
|
| 211 |
-
|
| 212 |
-
|
| 213 |
-
|
| 214 |
-
|
| 215 |
-
|
| 216 |
-
|
| 217 |
-
|
| 218 |
-
|
| 219 |
-
|
| 220 |
-
|
| 221 |
-
|
| 222 |
-
|
| 223 |
-
|
| 224 |
-
|
| 225 |
-
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
| 232 |
-
|
| 233 |
-
|
| 234 |
-
|
| 235 |
-
|
| 236 |
-
box-shadow: var(--apple-shadow-sm) !important;
|
| 237 |
-
}
|
| 238 |
-
|
| 239 |
-
.bot .message {
|
| 240 |
-
background: var(--apple-gray-2) !important;
|
| 241 |
-
color: var(--apple-text) !important;
|
| 242 |
-
border-radius: var(--apple-radius-lg) !important;
|
| 243 |
-
padding: 10px 16px !important;
|
| 244 |
-
box-shadow: var(--apple-shadow-sm) !important;
|
| 245 |
-
}
|
| 246 |
-
|
| 247 |
-
/* Buttons */
|
| 248 |
-
.gr-button {
|
| 249 |
-
font-family: -apple-system, BlinkMacSystemFont, 'SF Pro Display', sans-serif !important;
|
| 250 |
-
font-weight: 500 !important;
|
| 251 |
-
border-radius: var(--apple-radius-sm) !important;
|
| 252 |
-
transition: all 0.2s cubic-bezier(0.4, 0, 0.2, 1) !important;
|
| 253 |
-
border: none !important;
|
| 254 |
-
box-shadow: var(--apple-shadow-sm) !important;
|
| 255 |
-
}
|
| 256 |
-
|
| 257 |
-
.gr-button-primary, .gr-button-primary:hover {
|
| 258 |
-
background: var(--apple-blue) !important;
|
| 259 |
-
color: white !important;
|
| 260 |
-
font-size: 16px !important;
|
| 261 |
-
padding: 12px 24px !important;
|
| 262 |
-
letter-spacing: -0.01em !important;
|
| 263 |
-
}
|
| 264 |
-
|
| 265 |
-
.gr-button-primary:hover {
|
| 266 |
-
background: var(--apple-blue-hover) !important;
|
| 267 |
-
transform: translateY(-1px) !important;
|
| 268 |
-
box-shadow: var(--apple-shadow-md) !important;
|
| 269 |
-
}
|
| 270 |
-
|
| 271 |
-
.gr-button-primary:active {
|
| 272 |
-
transform: translateY(0) !important;
|
| 273 |
-
}
|
| 274 |
-
|
| 275 |
-
/* Video component */
|
| 276 |
-
.gr-video {
|
| 277 |
-
border-radius: var(--apple-radius) !important;
|
| 278 |
-
overflow: hidden !important;
|
| 279 |
-
box-shadow: var(--apple-shadow-lg) !important;
|
| 280 |
-
border: 1px solid var(--apple-gray-3) !important;
|
| 281 |
-
background: var(--apple-text) !important;
|
| 282 |
-
}
|
| 283 |
-
|
| 284 |
-
video {
|
| 285 |
-
border-radius: var(--apple-radius) !important;
|
| 286 |
-
}
|
| 287 |
-
|
| 288 |
-
/* Sidebar */
|
| 289 |
-
.gr-sidebar {
|
| 290 |
-
background: var(--apple-bg) !important;
|
| 291 |
-
border-radius: var(--apple-radius) !important;
|
| 292 |
-
border: 1px solid var(--apple-gray-3) !important;
|
| 293 |
-
box-shadow: var(--apple-shadow-md) !important;
|
| 294 |
-
padding: 20px !important;
|
| 295 |
-
}
|
| 296 |
-
|
| 297 |
-
/* Accordion */
|
| 298 |
-
.gr-accordion {
|
| 299 |
-
background: var(--apple-gray-2) !important;
|
| 300 |
-
border-radius: var(--apple-radius-sm) !important;
|
| 301 |
-
border: 1px solid var(--apple-gray-3) !important;
|
| 302 |
-
overflow: hidden !important;
|
| 303 |
-
margin-top: 16px !important;
|
| 304 |
-
}
|
| 305 |
-
|
| 306 |
-
.gr-accordion-header {
|
| 307 |
-
background: var(--apple-gray-2) !important;
|
| 308 |
-
padding: 12px 16px !important;
|
| 309 |
-
font-weight: 500 !important;
|
| 310 |
-
color: var(--apple-text) !important;
|
| 311 |
-
border: none !important;
|
| 312 |
-
}
|
| 313 |
-
|
| 314 |
-
.gr-accordion-header:hover {
|
| 315 |
-
background: var(--apple-gray-3) !important;
|
| 316 |
-
}
|
| 317 |
-
|
| 318 |
-
/* Gallery */
|
| 319 |
-
.gr-gallery {
|
| 320 |
-
background: transparent !important;
|
| 321 |
-
gap: 12px !important;
|
| 322 |
-
}
|
| 323 |
-
|
| 324 |
-
.gr-gallery-item {
|
| 325 |
-
border-radius: var(--apple-radius-sm) !important;
|
| 326 |
-
overflow: hidden !important;
|
| 327 |
-
box-shadow: var(--apple-shadow-sm) !important;
|
| 328 |
-
border: 1px solid var(--apple-gray-3) !important;
|
| 329 |
-
transition: all 0.2s cubic-bezier(0.4, 0, 0.2, 1) !important;
|
| 330 |
-
}
|
| 331 |
-
|
| 332 |
-
.gr-gallery-item:hover {
|
| 333 |
-
transform: scale(1.02) !important;
|
| 334 |
-
box-shadow: var(--apple-shadow-md) !important;
|
| 335 |
-
}
|
| 336 |
-
|
| 337 |
-
/* Progress bar */
|
| 338 |
-
.progress-bar {
|
| 339 |
-
background: var(--apple-gray-3) !important;
|
| 340 |
-
border-radius: 100px !important;
|
| 341 |
-
overflow: hidden !important;
|
| 342 |
-
height: 4px !important;
|
| 343 |
-
}
|
| 344 |
-
|
| 345 |
-
.progress-bar-fill {
|
| 346 |
-
background: var(--apple-blue) !important;
|
| 347 |
-
transition: width 0.3s cubic-bezier(0.4, 0, 0.2, 1) !important;
|
| 348 |
-
}
|
| 349 |
-
|
| 350 |
-
/* Labels */
|
| 351 |
-
label {
|
| 352 |
-
color: var(--apple-gray-5) !important;
|
| 353 |
-
font-size: 13px !important;
|
| 354 |
-
font-weight: 500 !important;
|
| 355 |
-
letter-spacing: -0.01em !important;
|
| 356 |
-
margin-bottom: 6px !important;
|
| 357 |
-
}
|
| 358 |
-
|
| 359 |
-
/* Inputs and form elements */
|
| 360 |
-
.gr-input, .gr-textbox, .gr-dropdown {
|
| 361 |
-
background: var(--apple-bg) !important;
|
| 362 |
-
border: 1px solid var(--apple-gray-4) !important;
|
| 363 |
-
border-radius: var(--apple-radius-sm) !important;
|
| 364 |
-
padding: 10px 12px !important;
|
| 365 |
-
font-size: 15px !important;
|
| 366 |
-
transition: all 0.2s cubic-bezier(0.4, 0, 0.2, 1) !important;
|
| 367 |
-
}
|
| 368 |
-
|
| 369 |
-
.gr-input:focus, .gr-textbox:focus, .gr-dropdown:focus {
|
| 370 |
-
border-color: var(--apple-blue) !important;
|
| 371 |
-
box-shadow: 0 0 0 3px rgba(0, 113, 227, 0.1) !important;
|
| 372 |
-
outline: none !important;
|
| 373 |
-
}
|
| 374 |
-
|
| 375 |
-
/* Markdown content */
|
| 376 |
-
.gr-markdown {
|
| 377 |
-
color: var(--apple-text) !important;
|
| 378 |
-
line-height: 1.6 !important;
|
| 379 |
-
}
|
| 380 |
-
|
| 381 |
-
.gr-markdown h2 {
|
| 382 |
-
font-size: 20px !important;
|
| 383 |
-
font-weight: 600 !important;
|
| 384 |
-
margin-bottom: 12px !important;
|
| 385 |
-
color: var(--apple-text) !important;
|
| 386 |
-
}
|
| 387 |
-
|
| 388 |
-
/* Loading states */
|
| 389 |
-
.gr-loading {
|
| 390 |
-
color: var(--apple-gray-5) !important;
|
| 391 |
-
}
|
| 392 |
-
|
| 393 |
-
/* Scrollbars */
|
| 394 |
-
::-webkit-scrollbar {
|
| 395 |
-
width: 8px;
|
| 396 |
-
height: 8px;
|
| 397 |
-
}
|
| 398 |
-
|
| 399 |
-
::-webkit-scrollbar-track {
|
| 400 |
-
background: var(--apple-gray-2);
|
| 401 |
-
border-radius: 10px;
|
| 402 |
-
}
|
| 403 |
-
|
| 404 |
-
::-webkit-scrollbar-thumb {
|
| 405 |
-
background: var(--apple-gray-4);
|
| 406 |
-
border-radius: 10px;
|
| 407 |
-
}
|
| 408 |
-
|
| 409 |
-
::-webkit-scrollbar-thumb:hover {
|
| 410 |
-
background: var(--apple-gray-5);
|
| 411 |
-
}
|
| 412 |
-
|
| 413 |
-
/* Animations */
|
| 414 |
-
@keyframes fadeIn {
|
| 415 |
-
from {
|
| 416 |
-
opacity: 0;
|
| 417 |
-
transform: translateY(10px);
|
| 418 |
-
}
|
| 419 |
-
to {
|
| 420 |
-
opacity: 1;
|
| 421 |
-
transform: translateY(0);
|
| 422 |
-
}
|
| 423 |
-
}
|
| 424 |
-
|
| 425 |
-
.message {
|
| 426 |
-
animation: fadeIn 0.3s cubic-bezier(0.4, 0, 0.2, 1);
|
| 427 |
-
}
|
| 428 |
-
|
| 429 |
-
/* Responsive adjustments */
|
| 430 |
-
@media (max-width: 768px) {
|
| 431 |
-
h1 {
|
| 432 |
-
font-size: 36px;
|
| 433 |
-
}
|
| 434 |
-
|
| 435 |
-
.contain {
|
| 436 |
-
padding: 16px;
|
| 437 |
-
}
|
| 438 |
-
|
| 439 |
-
.message {
|
| 440 |
-
max-width: 90% !important;
|
| 441 |
-
}
|
| 442 |
-
}
|
| 443 |
-
""") as demo:
|
| 444 |
gr.Markdown("# 🎬 FastVLM Video Captioning")
|
| 445 |
|
| 446 |
with gr.Row():
|
|
|
|
| 157 |
return video_summary, frame_previews
|
| 158 |
|
| 159 |
# Create the Gradio interface
|
| 160 |
+
# Create custom Apple-inspired theme
|
| 161 |
+
class AppleTheme(gr.themes.Base):
|
| 162 |
+
def __init__(self):
|
| 163 |
+
super().__init__(
|
| 164 |
+
primary_hue=gr.themes.colors.blue,
|
| 165 |
+
secondary_hue=gr.themes.colors.gray,
|
| 166 |
+
neutral_hue=gr.themes.colors.gray,
|
| 167 |
+
spacing_size=gr.themes.sizes.spacing_md,
|
| 168 |
+
radius_size=gr.themes.sizes.radius_md,
|
| 169 |
+
text_size=gr.themes.sizes.text_md,
|
| 170 |
+
font=[
|
| 171 |
+
gr.themes.GoogleFont("Inter"),
|
| 172 |
+
"-apple-system",
|
| 173 |
+
"BlinkMacSystemFont",
|
| 174 |
+
"SF Pro Display",
|
| 175 |
+
"SF Pro Text",
|
| 176 |
+
"Helvetica Neue",
|
| 177 |
+
"Helvetica",
|
| 178 |
+
"Arial",
|
| 179 |
+
"sans-serif"
|
| 180 |
+
],
|
| 181 |
+
font_mono=[
|
| 182 |
+
gr.themes.GoogleFont("SF Mono"),
|
| 183 |
+
"ui-monospace",
|
| 184 |
+
"Consolas",
|
| 185 |
+
"monospace"
|
| 186 |
+
]
|
| 187 |
+
)
|
| 188 |
+
super().set(
|
| 189 |
+
# Core colors
|
| 190 |
+
body_background_fill="*neutral_50",
|
| 191 |
+
body_background_fill_dark="*neutral_950",
|
| 192 |
+
button_primary_background_fill="*primary_500",
|
| 193 |
+
button_primary_background_fill_hover="*primary_600",
|
| 194 |
+
button_primary_text_color="white",
|
| 195 |
+
button_primary_border_color="*primary_500",
|
| 196 |
+
|
| 197 |
+
# Shadows
|
| 198 |
+
button_shadow="0 2px 8px rgba(0, 0, 0, 0.04)",
|
| 199 |
+
button_shadow_hover="0 4px 12px rgba(0, 0, 0, 0.08)",
|
| 200 |
+
block_shadow="0 4px 12px rgba(0, 0, 0, 0.08)",
|
| 201 |
+
|
| 202 |
+
# Border radius
|
| 203 |
+
button_large_radius="8px",
|
| 204 |
+
button_small_radius="6px",
|
| 205 |
+
block_radius="12px",
|
| 206 |
+
container_radius="12px",
|
| 207 |
+
|
| 208 |
+
# Borders
|
| 209 |
+
block_border_width="1px",
|
| 210 |
+
block_border_color="*neutral_200",
|
| 211 |
+
input_border_width="1px",
|
| 212 |
+
input_border_color="*neutral_300",
|
| 213 |
+
input_border_color_focus="*primary_500",
|
| 214 |
+
|
| 215 |
+
# Text
|
| 216 |
+
block_title_text_weight="600",
|
| 217 |
+
block_label_text_weight="500",
|
| 218 |
+
block_label_text_size="13px",
|
| 219 |
+
block_label_text_color="*neutral_600",
|
| 220 |
+
body_text_color="*neutral_900",
|
| 221 |
+
|
| 222 |
+
# Spacing
|
| 223 |
+
layout_gap="16px",
|
| 224 |
+
block_padding="20px",
|
| 225 |
+
|
| 226 |
+
# Specific components
|
| 227 |
+
chatbot_code_background_color="*neutral_100",
|
| 228 |
+
slider_color="*primary_500",
|
| 229 |
+
|
| 230 |
+
# Transitions
|
| 231 |
+
button_transition="all 0.2s cubic-bezier(0.4, 0, 0.2, 1)",
|
| 232 |
+
)
|
| 233 |
+
|
| 234 |
+
# Create the Gradio interface with the custom theme
|
| 235 |
+
with gr.Blocks(theme=AppleTheme()) as demo:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 236 |
gr.Markdown("# 🎬 FastVLM Video Captioning")
|
| 237 |
|
| 238 |
with gr.Row():
|