File size: 5,786 Bytes
5acd9c3
 
 
59e0f6a
5acd9c3
 
 
 
 
 
 
 
 
 
 
59e0f6a
 
 
5acd9c3
 
 
 
 
 
 
 
 
59e0f6a
 
 
 
5acd9c3
 
 
59e0f6a
 
 
 
5acd9c3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
83f2577
5acd9c3
 
 
 
 
 
 
 
 
 
83f2577
 
 
5acd9c3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
83f2577
 
 
5acd9c3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
83f2577
5acd9c3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
83f2577
5acd9c3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
83f2577
 
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
import os

PRODUCT_NAME = os.environ.get('PRODUCT_NAME', 'AiTube')
PRODUCT_VERSION = "2.0.0"

TEXT_MODEL = os.environ.get('HF_TEXT_MODEL',
    #'HuggingFaceH4/zephyr-7b-beta'
    'HuggingFaceTB/SmolLM2-1.7B-Instruct'
)

IMAGE_MODEL = os.environ.get('HF_IMAGE_MODEL', '')

# Environment variable to control maintenance mode
MAINTENANCE_MODE = os.environ.get('MAINTENANCE_MODE', 'false').lower() in ('true', 'yes', '1', 't')

# Environment variable to control how many nodes to use
MAX_NODES = int(os.environ.get('MAX_NODES', '8'))

ADMIN_ACCOUNTS = [
    "jbilcke-hf"
]

RAW_VIDEO_ROUND_ROBIN_ENDPOINT_URLS = [
    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_1', ''),
    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_2', ''),
    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_3', ''),
    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_4', ''),
    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_5', ''),
    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_6', ''),
    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_7', ''),
    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_8', ''),
]

# Filter out empty strings from the endpoint list
filtered_urls = [url for url in RAW_VIDEO_ROUND_ROBIN_ENDPOINT_URLS if url]

# Limit the number of URLs based on MAX_NODES environment variable
VIDEO_ROUND_ROBIN_ENDPOINT_URLS = filtered_urls[:MAX_NODES]

HF_TOKEN = os.environ.get('HF_TOKEN')

# use the same secret token as you used to secure your BASE_SPACE_NAME spaces
SECRET_TOKEN = os.environ.get('SECRET_TOKEN')

# altenative words we could use: "saturated, highlight, overexposed, highlighted, overlit, shaking, too bright, worst quality, inconsistent motion, blurry, jittery, distorted, cropped, watermarked, watermark, logo, subtitle, subtitles, lowres"
NEGATIVE_PROMPT = "low quality, worst quality, deformed, distorted, disfigured, blurry, text, watermark"

POSITIVE_PROMPT_SUFFIX = "high quality, cinematic, 4K, intricate details"

GUIDANCE_SCALE = 1.0

# anonymous users are people browing AiTube2 without being connected
# this category suffers from regular abuse so we need to enforce strict limitations
CONFIG_FOR_ANONYMOUS_USERS = {

    # anons can only watch 2 minutes per video
    "max_rendering_time_per_client_per_video_in_sec": 2 * 60,

    "max_buffer_size": 2,
    "max_concurrent_generations": 2,

    "min_num_inference_steps": 2,
    "default_num_inference_steps": 3,
    "max_num_inference_steps": 4,

    "min_num_frames": 9, # 8 + 1
    "default_max_num_frames": 65, # 8*8 + 1
    "max_num_frames": 65, # 8*8 + 1

    "min_clip_duration_seconds": 1,
    "default_clip_duration_seconds": 2,
    "max_clip_duration_seconds": 2,

    "min_clip_playback_speed": 0.7,
    "default_clip_playback_speed": 0.7,
    "max_clip_playback_speed": 0.7,

    "min_clip_framerate": 8,
    "default_clip_framerate": 16,
    "max_clip_framerate": 16,

    "min_clip_width": 544,
    "default_clip_width": 544,
    "max_clip_width": 544,

    "min_clip_height": 320,
    "default_clip_height": 320,
    "max_clip_height": 320,
}

# Hugging Face users enjoy a more normal and calibrated experience
CONFIG_FOR_STANDARD_HF_USERS = {
    "max_rendering_time_per_client_per_video_in_sec": 15 * 60,

    "max_buffer_size": 2,
    "max_concurrent_generations": 2,

    "min_num_inference_steps": 2,
    "default_num_inference_steps": 4,
    "max_num_inference_steps": 6,
    
    "min_num_frames": 9, # 8 + 1
    "default_num_frames": 65, # 8*8 + 1
    "max_num_frames": 65, # 8*8 + 1

    "min_clip_duration_seconds": 1,
    "default_clip_duration_seconds": 2,
    "max_clip_duration_seconds": 2,

    "min_clip_playback_speed": 0.7,
    "default_clip_playback_speed": 0.8,
    "max_clip_playback_speed": 0.8,

    "min_clip_framerate": 8,
    "default_clip_framerate": 25,
    "max_clip_framerate": 25,

    "min_clip_width": 544,
    "default_clip_width": 640,
    "max_clip_width": 640,

    "min_clip_height": 320,
    "default_clip_height": 416,
    "max_clip_height": 416,
}

# Hugging Face users with a Pro may enjoy an improved experience
CONFIG_FOR_PRO_HF_USERS = {
    "max_rendering_time_per_client_per_video_in_sec": 20 * 60,

    "max_buffer_size": 2,
    "max_concurrent_generations": 2,
    
    "min_num_inference_steps": 2,
    "max_num_inference_steps": 8,

    "min_num_frames": 9, # 8 + 1
    "default_num_frames": 97, # (8*12) + 1
    "max_num_frames": 97, # (8*12) + 1

    "min_clip_duration_seconds": 1, 
    "default_clip_duration_seconds": 2,
    "max_clip_duration_seconds": 3,

    "min_clip_playback_speed": 0.7,
    "default_clip_playback_speed": 0.8,
    "max_clip_playback_speed": 0.8,

    "min_clip_framerate": 8,
    "default_clip_framerate": 25,
    "max_clip_framerate": 30,

    "min_clip_width": 544,
    "default_clip_width": 768,
    "max_clip_width": 768,

    "min_clip_height": 320,
    "default_clip_height": 480,
    "max_clip_height": 480,
}

CONFIG_FOR_ADMIN_HF_USERS = {
    "max_rendering_time_per_client_per_video_in_sec": 60 * 60,

    "max_buffer_size": 2,
    "max_concurrent_generations": 2,
    
    "min_num_inference_steps": 2,
    "default_num_inference_steps": 4,
    "max_num_inference_steps": 8,

    "min_num_frames": 9, # 8 + 1
    "default_num_frames": 65, # (8 * 8) + 1
    "max_num_frames": 129, # (8 * 16) + 1

    "min_clip_duration_seconds": 1,
    "default_clip_duration_seconds": 2,
    "max_clip_duration_seconds": 4,

    "min_clip_playback_speed": 0.7,
    "default_clip_playback_speed": 0.8,
    "max_clip_playback_speed": 0.9,

    "min_clip_framerate": 8,
    "default_clip_framerate": 30,
    "max_clip_framerate": 60,

    "min_clip_width": 544,
    "default_clip_width": 768,
    "max_clip_width": 1216,

    "min_clip_height": 320,
    "default_clip_height": 480,
    "max_clip_height": 704,
}

CONFIG_FOR_ANONYMOUS_USERS = CONFIG_FOR_STANDARD_HF_USERS