-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathindex.html
executable file
·612 lines (578 loc) · 42.3 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
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
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
<!doctype html>
<html lang="en" xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<link rel="stylesheet" href="./assets/index.css">
<link rel='stylesheet' href='https://cdnjs.cloudflare.com/ajax/libs/twitter-bootstrap/4.6.2/css/bootstrap.min.css'>
<link rel="stylesheet" href="./assets/bootstrap.min.css">
<link rel="shortcut icon" href="data:image/svg+xml,<svg xmlns=%22http://www.w3.org/2000/svg%22 viewBox=%220 0 100 100%22><text y=%22.9em%22 font-size=%2290%22>⚗️</text></svg>">
<title>Video Alchemist</title>
</head>
<body data-new-gr-c-s-check-loaded="14.1110.0" data-gr-ext-installed="">
<div id="Video Alchemist" class="jumbotron text-center pb-2container" style="background-image: url('./data/background_smoke.jpg'); background-size: cover; background-position: 0% 5%;">
<br><br><br>
<p style="font-size: min(6rem, 10vw); margin-bottom: -10px; text-shadow: 0px 0px 10px #000;">⚗️ Video Alchemist</p>
<p class="responsive-text-title" style="font-size: 2rem; text-shadow: 0px 0px 10px #000;">Multi-subject Open-set Personalization in Video Generation</p>
<br>
<a href="https://tsaishien-chen.github.io/" class="responsive-text-context" style="color: white !important;"><b>Tsai-Shien Chen</b></a>,
<a href="https://aliaksandrsiarohin.github.io/aliaksandr-siarohin-website/" class="responsive-text-context" style="color: white !important;"><b>Aliaksandr Siarohin</b></a>,
<a href="https://www.willimenapace.com/" class="responsive-text-context" style="color: white !important;"><b>Willi Menapace</b></a>,
<a href="https://yuwfan.github.io/" class="responsive-text-context" style="color: white !important;"><b>Yuwei Fang</b></a>,
<a href="https://www.linkedin.com/in/kwotsin/" class="responsive-text-context" style="color: white !important;"><b>Kwot Sin Lee</b></a>,
<a href="https://skor.sh/" class="responsive-text-context" style="color: white !important;"><b>Ivan Skorokhodov</b></a>,
<br>
<a href="https://kfiraberman.github.io/" class="responsive-text-context" style="color: white !important;"><b>Kfir Aberman</b></a>,
<a href="https://www.cs.cmu.edu/~junyanz/" class="responsive-text-context" style="color: white !important;"><b>Jun-Yan Zhu</b></a>,
<a href="https://faculty.ucmerced.edu/mhyang/" class="responsive-text-context" style="color: white !important;"><b>Ming-Hsuan Yang</b></a>,
<a href="http://www.stulyakov.com/" class="responsive-text-context" style="color: white !important;"><b>Sergey Tulyakov</b></a>
<br><br>
<a href="https://research.snap.com/team/creative-vision"><img class="paper-btn" src="./data/logo_snap.svg" style="height: 80px; width: auto; background-color: #000;"></a>
<a href="https://www.ucmerced.edu/"><img class="paper-btn" src="./data/logo_ucmerced.svg" style="height: 80px; width: auto; background-color: #000;"></a>
<a href="https://www.cmu.edu/"><img class="paper-btn" src="./data/logo_cmu.svg" style="height: 80px; width: auto; background-color: #000;"></a>
<br><br>
<div class="paper-btn-parent">
<a class="paper-btn" style="width: 180px" href="https://arxiv.org/pdf/2501.06187">Read research paper</a>
<a class="paper-btn" style="width: 200px" href="./more_comparisons.html">View more comparisons</a>
</div>
</div>
<div class="container text-center">
<p class="responsive-text-title" style="font-size: 2rem;"><em>Video Alchemist</em> personalizes video generation</p>
<p class="responsive-text-context" style="margin-top: 15px; margin-bottom: 30px; max-width: 750px; margin-left: auto; margin-right: auto;">
Given a text prompt and a set of reference images conceptualizing entity words in the prompt, <em>Video Alchemist</em> generates the video conditioned on both text and reference images.
</p>
<div id="video-wrapper" style="position: relative; width: 100%; height: 0; padding-bottom: 56.25%; background: #2f2f30;">
<video id="video-id" src="./data/samples/ablate_background/1_sea_ice.mp4" playsinline autoplay loop muted style="position: absolute; top: 0; left: 0; width: 100%; height: 100%; border-radius: 2vmin; background: #2f2f30;"></video>
</div>
<div id="text-id" class="responsive-text-prompt" style="font-family: Chalkduster; margin-top:10px; margin-bottom:20px;">
"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."
</div>
<div id="carouselExampleCaptions-1" class="carousel slide" data-ride="carousel">
<div class="carousel-inner">
<div class="carousel-item active">
<div style="display: flex; align-items: center; justify-content: space-between; align-items: center;">
<a class="responsive-text-prompt custom-button" href="#carouselExampleCaptions-1" role="button" data-slide="prev" style="display: flex;">
<img src="./data/arrow-right.svg" style="width: min(50px, 5vw); height: min(50px, 5vw); transform: rotate(180deg);">
<div style="text-align: left; margin-left: 2%; margin-top: 5px; font-family: Chalkduster;">Switch<br><span style="color:#efc235;">background</span></div>
</a>
<div id="image-set-1" class="image-set">
<img src="./data/samples/ablate_person/0_no_image.svg" data-set="1" data-index="1" onclick="selectImage(this)">
<img src="./data/samples/ablate_person/1_yoshua_bengio.svg" data-set="1" data-index="2" onclick="selectImage(this)" class="selected">
<img src="./data/samples/ablate_person/2_geoffrey_hinton.svg" data-set="1" data-index="3" onclick="selectImage(this)">
<img src="./data/samples/ablate_person/3_yann_lecun.svg" data-set="1" data-index="4" onclick="selectImage(this)">
<img src="./data/samples/ablate_person/4_sora.svg" data-set="1" data-index="5" onclick="selectImage(this)">
<img src="./data/samples/ablate_person/5_dalle.svg" data-set="1" data-index="6" onclick="selectImage(this)">
</div>
<a class="responsive-text-prompt custom-button" href="#carouselExampleCaptions-1" role="button" data-slide="next" style="display: flex;">
<div style="text-align: right; margin-right: 2%; margin-top: 5px; font-family: Chalkduster;">Switch<br><span style="color:#68a74d;">dog</span></div>
<img src="./data/arrow-right.svg" style="width: min(50px, 5vw); height: min(50px, 5vw);">
</a>
</div>
</div>
<div class="carousel-item">
<div style="display: flex; align-items: center; justify-content: space-between; align-items: center;">
<a class="responsive-text-prompt custom-button" href="#carouselExampleCaptions-1" role="button" data-slide="prev" style="display: flex;">
<img src="./data/arrow-right.svg" style="width: min(50px, 5vw); height: min(50px, 5vw); transform: rotate(180deg);">
<div style="text-align: left; margin-left: 2%; margin-top: 5px; font-family: Chalkduster;">Switch<br><span style="color:#3e76d9;">person</span></div>
</a>
<div id="image-set-2" class="image-set">
<img src="./data/samples/ablate_dog/0_no_image.svg" data-set="2" data-index="1" onclick="selectImage(this)">
<img src="./data/samples/ablate_dog/1_shiba_inu.svg" data-set="2" data-index="2" onclick="selectImage(this)" class="selected">
<img src="./data/samples/ablate_dog/2_golden_retriever.svg" data-set="2" data-index="3" onclick="selectImage(this)">
<img src="./data/samples/ablate_dog/3_german_shepherd.png" data-set="2" data-index="4" onclick="selectImage(this)">
<img src="./data/samples/ablate_dog/4_husky.svg" data-set="2" data-index="5" onclick="selectImage(this)">
<img src="./data/samples/ablate_dog/5_samoyed.png" data-set="2" data-index="6" onclick="selectImage(this)">
</div>
<a class="responsive-text-prompt custom-button" href="#carouselExampleCaptions-1" role="button" data-slide="next" style="display: flex;">
<div style="text-align: right; margin-right: 2%; margin-top: 5px; font-family: Chalkduster;">Switch<br><span style="color:#efc235;">background</span></div>
<img src="./data/arrow-right.svg" style="width: min(50px, 5vw); height: min(50px, 5vw);">
</a>
</div>
</div>
<div class="carousel-item">
<div style="display: flex; align-items: center; justify-content: space-between; align-items: center;">
<a class="responsive-text-prompt custom-button" href="#carouselExampleCaptions-1" role="button" data-slide="prev" style="display: flex;">
<img src="./data/arrow-right.svg" style="width: min(50px, 5vw); height: min(50px, 5vw); transform: rotate(180deg);">
<div style="text-align: left; margin-left: 2%; margin-top: 5px; font-family: Chalkduster;">Switch<br><span style="color:#68a74d;">dog</span></div>
</a>
<div id="image-set-3" class="image-set">
<img src="./data/samples/ablate_background/0_no_image.svg" data-set="3" data-index="1" onclick="selectImage(this)">
<img src="./data/samples/ablate_background/1_sea_ice.svg" data-set="3" data-index="2" onclick="selectImage(this)" class="selected">
<img src="./data/samples/ablate_background/2_beach.svg" data-set="3" data-index="3" onclick="selectImage(this)">
<img src="./data/samples/ablate_background/3_london.svg" data-set="3" data-index="4" onclick="selectImage(this)">
<img src="./data/samples/ablate_background/4_desert.svg" data-set="3" data-index="5" onclick="selectImage(this)">
<img src="./data/samples/ablate_background/5_moon_surface.svg" data-set="3" data-index="6" onclick="selectImage(this)">
</div>
<a class="responsive-text-prompt custom-button" href="#carouselExampleCaptions-1" role="button" data-slide="next" style="display: flex;">
<div style="text-align: right; margin-right: 2%; margin-top: 5px; font-family: Chalkduster;">Switch<br><span style="color:#3e76d9;">person</span></div>
<img src="./data/arrow-right.svg" style="width: min(50px, 5vw); height: min(50px, 5vw);">
</a>
</div>
</div>
</div>
</div>
<p class="responsive-text-footnote" style="margin-top: 20px; margin-bottom: 50px; text-align: center;">
*Click the images to switch different reference images. Click the left and right buttons to switch the images of different conditional subjects: <em>person</em>, <em>dog</em>, or <em>background</em>.
</p>
</div>
<hr class="mt-5">
<div class="container text-center" style="margin-top: 50px;">
<p class="responsive-text-title"><em>Video Alchemist</em> supports multi-subject open-set personalization</p>
<p class="responsive-text-context" style="margin-top: 15px; max-width: 750px; margin-left: auto; margin-right: auto;">
<em>Video Alchemist</em> is equipped with built-in multi-subject, open-set personalization capabilities for both foreground objects and background, eliminating the need for test-time optimization.
</p>
</div>
<div id="carouselExampleCaptions" class="carousel slide show-neighbors" data-ride="carousel" data-interval="false">
<div class="carousel-inner">
<div class="carousel-item active">
<div class="item__third">
<div id="video-wrapper" style="position: relative; width: 100%; height: 0; padding-bottom: 56.25%; background: #2f2f30;">
<video id="video-set-1" src="./data/samples/multi_subject_dinosaur/3_conditions.mp4" playsinline autoplay loop muted style="position: absolute; top: 0; left: 0; width: 100%; height: 100%; border-radius: 2vmin; background: #2f2f30;"></video>
</div>
<div id="text-prompt-set-1" class="responsive-text-prompt" style="font-family: Chalkduster; color: white; margin-top:10px; margin-bottom:10px;">
"A <span style="color:#3e76d9;">woman</span> rides a <span style="color:#68a74d;">dinosaur</span> on a <span style="color:#efc235;">field</span>."
</div>
<div class="block-container mobile-break" style="align-items: center;">
<img src="./data/arrow-decrease.svg" class="custom-button" style="width: min(240px, 24vw);" onclick="changeCondition(1, -1);">
<div id="text-condition-set-1" class="responsive-text-context">3 conditional subjects</div>
<img src="./data/arrow-increase.svg" class="custom-button" style="width: min(240px, 24vw);" onclick="changeCondition(1, 1);">
</div>
</div>
</div>
<div class="carousel-item">
<div class="item__third">
<div id="video-wrapper" style="position: relative; width: 100%; height: 0; padding-bottom: 56.25%; background: #2f2f30;">
<video id="video-set-2" src="./data/samples/multi_subject_discuss/3_conditions.mp4" playsinline autoplay loop muted style="position: absolute; top: 0; left: 0; width: 100%; height: 100%; border-radius: 2vmin; background: #2f2f30;"></video>
</div>
<div id="text-prompt-set-2" class="responsive-text-prompt" style="font-family: Chalkduster; color: white; margin-top:10px; margin-bottom:10px;">
"A <span style="color:#3e76d9;">man</span> and a <span style="color:#68a74d;">woman</span> discuss something in a <span style="color:#efc235;">meeting room</span>."
</div>
<div class="block-container mobile-break" style="align-items: center;">
<img src="./data/arrow-decrease.svg" class="custom-button" style="width: min(240px, 24vw);" onclick="changeCondition(2, -1);">
<div id="text-condition-set-2" class="responsive-text-context">3 conditional subjects</div>
<img src="./data/arrow-increase.svg" class="custom-button" style="width: min(240px, 24vw);" onclick="changeCondition(2, 1);">
</div>
</div>
</div>
<div class="carousel-item">
<div class="item__third">
<div id="video-wrapper" style="position: relative; width: 100%; height: 0; padding-bottom: 56.25%; background: #2f2f30;">
<video id="video-set-3" src="./data/samples/multi_subject_drink_tea/3_conditions.mp4" playsinline autoplay loop muted style="position: absolute; top: 0; left: 0; width: 100%; height: 100%; border-radius: 2vmin; background: #2f2f30;"></video>
</div>
<div id="text-prompt-set-3" class="responsive-text-prompt" style="font-family: Chalkduster; color: white; margin-top:10px; margin-bottom:10px;">
"A <span style="color:#3e76d9;">woman</span> in a <span style="color:#68a74d;">suit</span> sits in a <span style="color:#efc235;">living room</span> and drinks tea."
</div>
<div class="block-container mobile-break" style="align-items: center;">
<img src="./data/arrow-decrease.svg" class="custom-button" style="width: min(240px, 24vw);" onclick="changeCondition(3, -1);">
<div id="text-condition-set-3" class="responsive-text-context">3 conditional subjects</div>
<img src="./data/arrow-increase.svg" class="custom-button" style="width: min(240px, 24vw);" onclick="changeCondition(3, 1);">
</div>
</div>
</div>
<div class="carousel-item">
<div class="item__third">
<div id="video-wrapper" style="position: relative; width: 100%; height: 0; padding-bottom: 56.25%; background: #2f2f30;">
<video id="video-set-4" src="./data/samples/multi_subject_rocket/3_conditions.mp4" playsinline autoplay loop muted style="position: absolute; top: 0; left: 0; width: 100%; height: 100%; border-radius: 2vmin; background: #2f2f30;"></video>
</div>
<div id="text-prompt-set-4" class="responsive-text-prompt" style="font-family: Chalkduster; color: white; margin-top:10px; margin-bottom:10px;">
"A <span style="color:#3e76d9;">rocket</span> launches from the <span style="color:#efc235;">Moon’s surface</span> with a <span style="color:#68a74d;">UFO</span> behind."
</div>
<div class="block-container mobile-break" style="align-items: center;">
<img src="./data/arrow-decrease.svg" class="custom-button" style="width: min(240px, 24vw);" onclick="changeCondition(4, -1);">
<div id="text-condition-set-4" class="responsive-text-context">3 conditional subjects</div>
<img src="./data/arrow-increase.svg" class="custom-button" style="width: min(240px, 24vw);" onclick="changeCondition(4, 1);">
</div>
</div>
</div>
</div>
<a class="carousel-control-prev" href="#carouselExampleCaptions" role="button" data-slide="prev">
<span class="carousel-control-prev-icon" aria-hidden="true" style="background-image: url('./data/arrow-right.svg'); width: 50px; height: 50px; transform: rotate(180deg);"></span>
<span class="sr-only">Previous</span>
</a>
<a class="carousel-control-next" href="#carouselExampleCaptions" role="button" data-slide="next">
<span class="carousel-control-next-icon" aria-hidden="true" style="background-image: url('./data/arrow-right.svg'); width: 50px; height: 50px;"></span>
<span class="sr-only">Next</span>
</a>
<p class="responsive-text-footnote" style="margin-bottom: 50px; text-align: center;">
*Click the up and down buttons to change the number of conditional subjects.
</p>
</div>
<hr class="mt-5">
<div class="container text-center" style="margin-top: 50px;">
<p class="responsive-text-title"><em>Video Alchemist</em> is the state-of-the-art personalization model</p>
<p class="responsive-text-context" style="margin-top: 15px; margin-bottom: 30px; max-width: 725px; margin-left: auto; margin-right: auto;">
Compared to the existing personalization models, <em>Video Alchemist</em> can generate videos with the best text alignment, the highest subject fidelity, and the largest video dynamic<sup>†</sup>.
</p>
<div class="block-container" style="margin-bottom: 10px; gap: 6px;">
<img src="./data/samples/msrvtt_horse/inputs.svg" style="width: 100%;">
<div style="width: 15.9%; min-width: 106px;">
<video src="./data/samples/msrvtt_horse/0_ELITE.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">ELITE</p>
</div>
<div style="width: 15.9%; min-width: 106px;">
<video src="./data/samples/msrvtt_horse/1_VideoBooth.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">VideoBooth</p>
</div>
<div style="width: 15.9%; min-width: 106px;">
<video src="./data/samples/msrvtt_horse/2_DreamVideo.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">DreamVideo</p>
</div>
<div style="width: 28.25%; min-width: 188.3px;">
<video src="./data/samples/msrvtt_horse/3_Video_Alchemist.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context"><em>Video Alchemist</em></p>
</div>
<div style="width: 21.2%; min-width: 141.3px;">
<video src="./data/samples/msrvtt_horse/4_Ground_Truth.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">Ground Truth</p>
</div>
</div>
<div class="block-container" style="margin-bottom: 10px; gap: 6px;">
<img src="./data/samples/msrvtt_face_0/inputs.svg" style="width: 100%;">
<div style="width: 23.62%; min-width: 157.5px;">
<video src="./data/samples/msrvtt_face_0/0_IP-Adapter.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">IP-Adapter</p>
</div>
<div style="width: 13.3%; min-width: 88.67px;">
<video src="./data/samples/msrvtt_face_0/1_PhotoMaker.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">PhotoMaker</p>
</div>
<div style="width: 13.3%; min-width: 88.67px;">
<video src="./data/samples/msrvtt_face_0/2_Magic-Me.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">Magic-Me</p>
</div>
<div style="width: 23.62%; min-width: 157.5px;">
<video src="./data/samples/msrvtt_face_0/3_Video_Alchemist.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context"><em>Video Alchemist</em></p>
</div>
<div style="width: 23.62%; min-width: 157.5px;">
<video src="./data/samples/msrvtt_face_0/4_Ground_Truth.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">Ground Truth</p>
</div>
</div>
<div style="display: flex; justify-content: space-around; flex-wrap: wrap;">
<a class="paper-btn" style="width: 200px" href="./more_comparisons.html">View more comparisons</a>
</div>
<p class="responsive-text-footnote" style="margin-top: 25px; margin-bottom: 50px; text-align: center;">
<sup>†</sup>According to the evaluation on the proposed benchmark <em>MSRVTT-Personalization</em>. Check the following section or the research paper for more details.
</p>
</div>
<hr class="mt-5">
<div class="container text-center" style="margin-top: 50px;">
<p class="responsive-text-title">How do we make <em>Video Alchemist</em> work?</p>
<p class="responsive-text-context" style="margin-top: 15px; margin-bottom: 25px; max-width: 950px; margin-left: auto; margin-right: auto;">
To achieve multi-subject, open-set personalization, we design a dataset construction pipeline, including retrieving the entity words and preparing the subject and background images.
To mitigate the issue of "<em>copy-and-paste</em>" effect, we collect subject images from multiple frames and introduce training-time image augmentations.
</p>
<video src="./data/video_dataset_collection.mp4" playsinline autoplay loop muted controls style="width: 100%; border-radius: 2vmin; background: #2f2f30;"></video>
<p class="responsive-text-context" style="margin-top: 50px; margin-bottom: 25px; max-width: 850px; margin-left: auto; margin-right: auto;">
<em>Video Alchemist</em> is built on new Diffusion Transformer modules with additional cross-attention layer for personalization conditioning.
To achieve multi-subject conditioning, we introduce a subject-level fusion, binding the word description of each subject with its image representations.
</p>
<video src="./data/video_model_architecture.mp4" playsinline autoplay loop muted controls style="width: 100%; border-radius: 2vmin; background: #2f2f30;"></video>
<div style="display: flex; justify-content: space-around; flex-wrap: wrap; margin-top: 30px;">
<a class="paper-btn" style="width: 180px; margin-bottom: 50px;" href="https://arxiv.org/pdf/2501.06187">Read research paper</a>
</div>
</div>
<hr class="mt-5">
<div class="container text-center" style="margin-top: 50px;">
<p class="responsive-text-title"><em>MSRVTT-Personalization</em>: a new benchmark of personalization</p>
<p class="responsive-text-context" style="margin-top: 15px; margin-bottom: 30px; max-width: 1000px; margin-left: auto; margin-right: auto;">
To evaluate <em>Video Alchemist</em>, we introduce <em>MSRVTT-Personalization</em> aiming at accurate subject fidelity assessment and supporting various conditioning modes, including conditioning on face crops, single or multiple arbitrary subjects, and the combination of foreground objects and background. We show a test sample in <em>MSRVTT-Personalization</em> below.
</p>
<div class="block-container" style="flex-wrap: none; gap: .5%; margin-bottom: 5px;">
<div style="width: 29%;">
<video src="./data/msrvtt_sample/video.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">Ground Truth Video</p>
</div>
<div style="width: 48%;">
<img src="./data/msrvtt_sample/annotation.png" style="width: 100%;">
<p class="responsive-text-context">Personalization Annotations</p>
</div>
<div style="width: 19.1%;">
<img src="./data/msrvtt_sample/metrics.svg" style="width: 100%;">
<p class="responsive-text-context">Evaluation Metrics</p>
</div>
</div>
<div style="display: flex; justify-content: space-around; flex-wrap: wrap;">
<a class="paper-btn" style="width: 250px; margin-bottom: 50px;" href="https://github.com/snap-research/MSRVTT-Personalization">Benchmark dataset and code</a>
</div>
</div>
<hr class="mt-5">
<div class="container text-center" style="margin-top: 50px;">
<p class="responsive-text-title">Ablation study</p>
<p class="responsive-text-context" style="margin-top: 15px; margin-bottom: 30px;">
<em>Video Alchemist</em> can achieve better subject fidelity using DINOv2 as the encoder compared to CLIP <sup>[1]</sup>, <br>
it can correctly bind the reference image and the corresponding entity word with the usage of word tokens <sup>[2]</sup>, <br>
and it can mitigate the <em>copy-and-paste</em> effect and synthesize text-aligned videos via the proposed image augmentations <sup>[3]</sup>.
</p>
<div class="block-container" style="margin-bottom: 30px; gap: 6px;">
<img src="./data/samples/ablation_study/inputs.svg" style="width: 100%;">
<div style="width: 24.5%; min-width: 200px;">
<video src="./data/samples/ablation_study/0_use_clip.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">Use CLIP <sup>[1]</sup></p>
</div>
<div style="width: 24.5%; min-width: 200px;">
<video src="./data/samples/ablation_study/1_no_word_token.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">No word token <sup>[2]</sup></p>
</div>
<div style="width: 24.5%; min-width: 200px;">
<video src="./data/samples/ablation_study/2_no_augmentation.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context">No augmentation <sup>[3]</sup></p>
</div>
<div style="width: 24.5%; min-width: 200px;">
<video src="./data/samples/ablation_study/3_Video_Alchemist.mp4" loop playsinline autoPlay muted style="width: 100%; border-radius: 1vmin;"></video>
<p class="responsive-text-context"><em>Video Alchemist</em></p>
</div>
</div>
</div>
<hr class="mt-5">
<div class="container text-center" style="margin-top: 50px; margin-bottom: 150px;">
<p class="responsive-text-title">Acknowledgement</p>
<p class="responsive-text-context text-left" style="margin-top: 15px; margin-bottom: 30px;">
We thank Ziyi Wu, Moayed Haji Ali, and Alper Canberk for their helpful discussions,
and also extend our gratitude to Snap Inc. for providing the computational resources and fostering a conducive research environment. 🤗 🙏 👻
</p>
</div>
<script>
// Set the default selected values for each set
const selectedImages = { 1: 2, 2: 2, 3: 2 };
// Handles image selection
function selectImage(img) {
const set = img.getAttribute("data-set");
const index = img.getAttribute("data-index");
// Deselect all images in the current set
const images = document.querySelectorAll(`[data-set="${set}"]`);
images.forEach(image => image.classList.remove("selected"));
// Select the clicked image
img.classList.add("selected");
// Update the selectedImages object for the current set
selectedImages[set] = index;
// Reset the other sets to the default value of 2
Object.keys(selectedImages).forEach(key => {
if (key !== set) {
selectedImages[key] = 2;
// Reset selection visually for other sets
const otherImages = document.querySelectorAll(`[data-set="${key}"]`);
otherImages.forEach(image => image.classList.remove("selected"));
const defaultImage = document.querySelector(`[data-set="${key}"][data-index="2"]`);
if (defaultImage) {
defaultImage.classList.add("selected");
}
}
});
// Update the video based on the new selections
const videoWrapper = document.getElementById("video-wrapper");
const videoElement = document.getElementById("video-id");
const textElement = document.getElementById("text-id");
if (selectedImages[1] == 1 && selectedImages[2] == 2 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_person/0_no_image.mp4`;
textElement.innerHTML = `"A man pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 2 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_person/1_yoshua_bengio.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 3 && selectedImages[2] == 2 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_person/2_geoffrey_hinton.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 4 && selectedImages[2] == 2 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_person/3_yann_lecun.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 5 && selectedImages[2] == 2 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_person/4_sora.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">woman</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 6 && selectedImages[2] == 2 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_person/5_dalle.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">woman</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 1 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_dog/0_no_image.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a dog on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 2 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_dog/1_shiba_inu.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 3 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_dog/2_golden_retriever.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 4 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_dog/3_german_shepherd.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 5 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_dog/4_husky.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 6 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_dog/5_samoyed.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 1 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_background/1_sea_ice.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 2 && selectedImages[3] == 1) {
videoElement.src = `./data/samples/ablate_background/0_no_image.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 2 && selectedImages[3] == 2) {
videoElement.src = `./data/samples/ablate_background/1_sea_ice.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">sea ice</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 2 && selectedImages[3] == 3) {
videoElement.src = `./data/samples/ablate_background/2_beach.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">beach</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 2 && selectedImages[3] == 4) {
videoElement.src = `./data/samples/ablate_background/3_london.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">bridge</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 2 && selectedImages[3] == 5) {
videoElement.src = `./data/samples/ablate_background/4_desert.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">desert</span>."`;
} else if (selectedImages[1] == 2 && selectedImages[2] == 2 && selectedImages[3] == 6) {
videoElement.src = `./data/samples/ablate_background/5_moon_surface.mp4`;
textElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> pets a <span style="color:#68a74d;">dog</span> on <span style="color:#efc235;">Moon's surface</span>."`;
}
// Update video source and load
videoElement.load();
videoElement.play();
}
// Track the current condition for each sample
var currentConditions = { 1: 3, 2: 3, 3: 3, 4: 3 };
function changeCondition(sampleId, change) {
// Update currentConditions
change = (change === -1) ? 3 : change;
currentConditions[sampleId] = (currentConditions[sampleId] + change) % 4;
// Update video and text source
const textConditionElements = document.querySelectorAll(`#text-condition-set-${sampleId}`);
textConditionElements.forEach((textConditionElement) => {
if (textConditionElement) {
textConditionElement.innerHTML = `${currentConditions[sampleId]} conditional subjects`;
}
});
const videoElements = document.querySelectorAll(`#video-set-${sampleId}`);
videoElements.forEach((videoElement) => {
if (videoElement) {
videoElement.src = `./data/samples/multi_subject_${getSampleType(parseInt(sampleId))}/${currentConditions[sampleId]}_conditions.mp4`;
videoElement.load();
videoElement.play();
}
});
const textPromptElements = document.querySelectorAll(`#text-prompt-set-${sampleId}`);
textPromptElements.forEach((textPromptElement) => {
if (parseInt(sampleId) == 1 && currentConditions[sampleId] == 0) {
textPromptElement.innerHTML = `"A woman rides a dinosaur on a field."`;
} else if (parseInt(sampleId) == 1 && currentConditions[sampleId] == 1) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">woman</span> rides an animal."`;
} else if (parseInt(sampleId) == 1 && currentConditions[sampleId] == 2) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">woman</span> rides a <span style="color:#68a74d;">dinosaur</span>."`;
} else if (parseInt(sampleId) == 1 && currentConditions[sampleId] == 3) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">woman</span> rides a <span style="color:#68a74d;">dinosaur</span> on a <span style="color:#efc235;">field</span>."`;
} else if (parseInt(sampleId) == 2 && currentConditions[sampleId] == 0) {
textPromptElement.innerHTML = `"A man and a woman discuss something in a meeting room."`;
} else if (parseInt(sampleId) == 2 && currentConditions[sampleId] == 1) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> and a woman discuss something."`;
} else if (parseInt(sampleId) == 2 && currentConditions[sampleId] == 2) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> and a <span style="color:#68a74d;">woman</span> discuss something."`;
} else if (parseInt(sampleId) == 2 && currentConditions[sampleId] == 3) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">man</span> and a <span style="color:#68a74d;">woman</span> discuss something in a <span style="color:#efc235;">meeting room</span>."`;
} else if (parseInt(sampleId) == 2 && currentConditions[sampleId] == 0) {
textPromptElement.innerHTML = `"A woman in a suit sits in a living room and drinks tea."`;;
} else if (parseInt(sampleId) == 3 && currentConditions[sampleId] == 0) {
textPromptElement.innerHTML = `"A woman in a suit sits in a living room and drinks tea."`;
} else if (parseInt(sampleId) == 3 && currentConditions[sampleId] == 1) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">woman</span> in a suit drinks tea."`;
} else if (parseInt(sampleId) == 3 && currentConditions[sampleId] == 2) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">woman</span> in a <span style="color:#68a74d;">suit</span> drinks tea."`;
} else if (parseInt(sampleId) == 3 && currentConditions[sampleId] == 3) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">woman</span> in a <span style="color:#68a74d;">suit</span> sits in a <span style="color:#efc235;">living room</span> and drinks tea."`;
} else if (parseInt(sampleId) == 4 && currentConditions[sampleId] == 0) {
textPromptElement.innerHTML = `"A rocket launches from the Moon’s surface with a UFO behind."`;
} else if (parseInt(sampleId) == 4 && currentConditions[sampleId] == 1) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">rocket</span> launches."`;
} else if (parseInt(sampleId) == 4 && currentConditions[sampleId] == 2) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">rocket</span> launches from the <span style="color:#efc235;">Moon’s surface</span>."`;
} else if (parseInt(sampleId) == 4 && currentConditions[sampleId] == 3) {
textPromptElement.innerHTML = `"A <span style="color:#3e76d9;">rocket</span> launches from the <span style="color:#efc235;">Moon’s surface</span> with a <span style="color:#68a74d;">UFO</span> behind."`;
}
});
}
function getSampleType(sampleId) {
// Map sample IDs to their respective video folder types
switch (sampleId) {
case 1: return "dinosaur";
case 2: return "discuss";
case 3: return "drink_tea";
case 4: return "rocket";
default: return "";
}
}
function updateFontSizeForClass(className) {
var elements = document.getElementsByClassName(className);
var screenWidth = window.innerWidth || document.documentElement.clientWidth || document.body.clientWidth;
if (className == "responsive-text-title") {
for (var i = 0; i < elements.length; i++) {
if (screenWidth < 600) {
elements[i].style.fontSize = "1.55rem";
} else if (screenWidth < 800) {
elements[i].style.fontSize = "1.7rem";
} else if (screenWidth < 1000) {
elements[i].style.fontSize = "1.85rem";
} else {
elements[i].style.fontSize = "2.0rem";
}
}
}
else if (className == "responsive-text-context") {
for (var i = 0; i < elements.length; i++) {
elements[i].style.fontWeight = "300";
if (screenWidth < 600) {
elements[i].style.fontSize = "13.5px";
} else if (screenWidth < 800) {
elements[i].style.fontSize = "15px";
} else if (screenWidth < 1000) {
elements[i].style.fontSize = "16.5px";
} else {
elements[i].style.fontSize = "18px";
}
}
}
else if (className == "responsive-text-prompt") {
for (var i = 0; i < elements.length; i++) {
// Adjust the threshold and font size as needed
if (screenWidth < 600) {
elements[i].style.fontSize = "9.5px";
} else if (screenWidth < 800) {
elements[i].style.fontSize = "11px";
} else if (screenWidth < 1000) {
elements[i].style.fontSize = "12.5px";
} else {
elements[i].style.fontSize = "14px";
}
}
}
else if (className == "responsive-text-footnote") {
for (var i = 0; i < elements.length; i++) {
// Adjust the threshold and font size as needed
if (screenWidth < 600) {
elements[i].style.fontSize = "8.5px";
} else if (screenWidth < 800) {
elements[i].style.fontSize = "10px";
} else if (screenWidth < 1000) {
elements[i].style.fontSize = "11.5px";
} else {
elements[i].style.fontSize = "13px";
}
elements[i].style.fontWeight = "200";
}
}
}
// Call the function on page load and window resize
window.onload = function() {
updateFontSizeForClass("responsive-text-title");
updateFontSizeForClass("responsive-text-context");
updateFontSizeForClass("responsive-text-prompt");
updateFontSizeForClass("responsive-text-footnote");
};
window.onresize = function() {
updateFontSizeForClass("responsive-text-title");
updateFontSizeForClass("responsive-text-context");
updateFontSizeForClass("responsive-text-prompt");
updateFontSizeForClass("responsive-text-footnote");
};
</script>
<script src='https://cdnjs.cloudflare.com/ajax/libs/jquery/3.7.1/jquery.min.js'></script>
<script src='https://cdnjs.cloudflare.com/ajax/libs/twitter-bootstrap/4.6.2/js/bootstrap.min.js'></script>
<script src="./assets/index.js"></script>
</body>
</html>