ginipick commited on
Commit
ee19fd5
Β·
verified Β·
1 Parent(s): 0345ed6

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +95 -1
README.md CHANGED
@@ -10,5 +10,99 @@ pinned: false
10
  license: mit
11
  short_description: input text, a video from the past to the future
12
  ---
 
13
 
14
- Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  license: mit
11
  short_description: input text, a video from the past to the future
12
  ---
13
+ Looking at this code, it's a Gradio-based application that generates interpolated images between two concepts using CLIP-guided diffusion with the FLUX model. Let me explain the key components and functionality:
14
 
15
+ ## English Explanation
16
+
17
+ ### Overview
18
+ This application creates a "Time Stream" effect by generating a series of images that smoothly transition between two different states or concepts. For example, it can show the progression from a "fresh" tomato to a "rotten" one, creating a time-lapse-like visualization.
19
+
20
+ ### Key Features
21
+
22
+ 1. **CLIP-Guided Image Generation**
23
+ - Uses FLUX.1-schnell model with CLIP guidance
24
+ - Finds latent directions between two concepts using CLIP embeddings
25
+ - Generates intermediate images along this direction
26
+
27
+ 2. **Main Components**
28
+ - **Prompt**: The base description of what to generate
29
+ - **1st/2nd Direction**: Two states to interpolate between (e.g., "Fresh" β†’ "Rotten")
30
+ - **Strength**: Controls how extreme the transformation is
31
+ - **Output**: Creates both an image strip and a looping video
32
+
33
+ 3. **Advanced Options**
34
+ - Number of intermediate images (3-65)
35
+ - CLIP direction iterations (0-400)
36
+ - Inference steps (1-4)
37
+ - Guidance scale (0.1-10.0)
38
+ - Seed control for reproducibility
39
+
40
+ 4. **Output Formats**
41
+ - Individual generated images
42
+ - Image strip showing all transitions
43
+ - Looping video of the transformation
44
+ - Interactive slider to view specific frames
45
+
46
+ ### Technical Implementation
47
+ - Uses `spaces.GPU` decorator for GPU acceleration
48
+ - Implements AutoencoderTiny for faster processing
49
+ - Handles Korean text detection (though warns it's used directly without translation)
50
+ - Saves images with unique UUID filenames
51
+
52
+ ### Example Use Cases
53
+ - Showing decay/aging processes
54
+ - Seasonal changes
55
+ - Weather transitions
56
+ - Urban development/deterioration
57
+ - Any temporal transformation
58
+
59
+ ---
60
+
61
+ ## ν•œκΈ€ μ„€λͺ…
62
+
63
+ ### κ°œμš”
64
+ 이 μ• ν”Œλ¦¬μΌ€μ΄μ…˜μ€ 두 κ°€μ§€ λ‹€λ₯Έ μƒνƒœλ‚˜ κ°œλ… 사이λ₯Ό λΆ€λ“œλŸ½κ²Œ μ „ν™˜ν•˜λŠ” 일련의 이미지λ₯Ό μƒμ„±ν•˜μ—¬ "μ‹œκ°„μ˜ 흐름(Time Stream)" 효과λ₯Ό λ§Œλ“­λ‹ˆλ‹€. 예λ₯Ό λ“€μ–΄, "μ‹ μ„ ν•œ" ν† λ§ˆν† μ—μ„œ "썩은" ν† λ§ˆν† λ‘œμ˜ λ³€ν™” 과정을 λ³΄μ—¬μ£ΌλŠ” μ‹œκ°„ κ²½κ³Ό μ‹œκ°ν™”λ₯Ό 생성할 수 μžˆμŠ΅λ‹ˆλ‹€.
65
+
66
+ ### μ£Όμš” κΈ°λŠ₯
67
+
68
+ 1. **CLIP κ°€μ΄λ“œ 이미지 생성**
69
+ - CLIP κ°€μ΄λ˜μŠ€μ™€ ν•¨κ»˜ FLUX.1-schnell λͺ¨λΈ μ‚¬μš©
70
+ - CLIP μž„λ² λ”©μ„ μ‚¬μš©ν•˜μ—¬ 두 κ°œλ… μ‚¬μ΄μ˜ 잠재 λ°©ν–₯ μ°ΎκΈ°
71
+ - 이 λ°©ν–₯을 따라 쀑간 이미지듀을 생성
72
+
73
+ 2. **μ£Όμš” ꡬ성 μš”μ†Œ**
74
+ - **ν”„λ‘¬ν”„νŠΈ**: 생성할 λŒ€μƒμ˜ κΈ°λ³Έ μ„€λͺ…
75
+ - **1μ°¨/2μ°¨ λ°©ν–₯**: 보간할 두 κ°€μ§€ μƒνƒœ (예: "μ‹ μ„ ν•œ" β†’ "썩은")
76
+ - **강도**: λ³€ν™˜μ˜ 극단성을 μ œμ–΄
77
+ - **좜λ ₯**: 이미지 슀트립과 루핑 λΉ„λ””μ˜€ λͺ¨λ‘ 생성
78
+
79
+ 3. **κ³ κΈ‰ μ˜΅μ…˜**
80
+ - 쀑간 이미지 수 (3-65개)
81
+ - CLIP λ°©ν–₯ 반볡 횟수 (0-400회)
82
+ - μΆ”λ‘  단계 (1-4단계)
83
+ - κ°€μ΄λ˜μŠ€ μŠ€μΌ€μΌ (0.1-10.0)
84
+ - μž¬ν˜„μ„±μ„ μœ„ν•œ μ‹œλ“œ μ œμ–΄
85
+
86
+ 4. **좜λ ₯ ν˜•μ‹**
87
+ - κ°œλ³„ 생성 이미지
88
+ - λͺ¨λ“  μ „ν™˜μ„ λ³΄μ—¬μ£ΌλŠ” 이미지 슀트립
89
+ - λ³€ν™˜ κ³Όμ •μ˜ 루핑 λΉ„λ””μ˜€
90
+ - νŠΉμ • ν”„λ ˆμž„μ„ λ³Ό 수 μžˆλŠ” μΈν„°λž™ν‹°λΈŒ μŠ¬λΌμ΄λ”
91
+
92
+ ### 기술적 κ΅¬ν˜„
93
+ - GPU 가속을 μœ„ν•œ `spaces.GPU` λ°μ½”λ ˆμ΄ν„° μ‚¬μš©
94
+ - λΉ λ₯Έ 처리λ₯Ό μœ„ν•œ AutoencoderTiny κ΅¬ν˜„
95
+ - ν•œκΈ€ ν…μŠ€νŠΈ 감지 처리 (λ²ˆμ—­ 없이 직접 μ‚¬μš©λœλ‹€λŠ” κ²½κ³  ν‘œμ‹œ)
96
+ - κ³ μœ ν•œ UUID 파일λͺ…μœΌλ‘œ 이미지 μ €μž₯
97
+
98
+ ### μ‚¬μš© μ˜ˆμ‹œ
99
+ - λΆ€νŒ¨/λ…Έν™” κ³Όμ • ν‘œν˜„
100
+ - κ³„μ ˆ λ³€ν™”
101
+ - 날씨 μ „ν™˜
102
+ - λ„μ‹œ 개발/쇠퇴
103
+ - λͺ¨λ“  μ‹œκ°„μ  λ³€ν™˜
104
+
105
+ ### 참고사항
106
+ - ν•œκΈ€ μž…λ ₯은 μ§€μ›λ˜μ§€λ§Œ λͺ¨λΈμ΄ μ˜μ–΄μ— μ΅œμ ν™”λ˜μ–΄ μžˆμ–΄ κ²°κ³Όκ°€ μ œν•œμ μΌ 수 있음
107
+ - 강도(Strength) 값이 2.5 이상일 경우 λΆˆμ•ˆμ •ν•  수 있음
108
+ - 쀑간 이미지 μˆ˜κ°€ λ§Žμ„μˆ˜λ‘ 더 λΆ€λ“œλŸ¬μš΄ μ „ν™˜ 효과λ₯Ό 얻을 수 있음