Sentinel7 commited on
Commit
61a9005
·
verified ·
1 Parent(s): ceea3cf

Upload 1939453/2585269/README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. 1939453/2585269/README.md +182 -3
1939453/2585269/README.md CHANGED
@@ -1,10 +1,189 @@
1
  ---
2
  license: other
3
  tags:
4
- - civitai
5
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  Author: [xiaozhijason](https://civitai.com/user/xiaozhijason)
7
 
8
- Model: [https://civitai.com/models/1939453?modelVersionId=2585269](https://civitai.com/models/1939453?modelVersionId=2585269)
 
 
9
 
10
- Mirror: [https://civarchive.com/models/1939453?modelVersionId=2585269](https://civarchive.com/models/1939453?modelVersionId=2585269)
 
1
  ---
2
  license: other
3
  tags:
4
+ - tool
5
  ---
6
+
7
+ # Consistence Edit Lora - qe2511 alpha - 2585269
8
+
9
+ **Model Type**: LORA
10
+
11
+ **Base Model**: Qwen
12
+
13
+ **Trigger Words**: None
14
+
15
+ **Tags**: tool
16
+
17
+ ## Gallery
18
+
19
+ <table>
20
+ <tr>
21
+ <td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ea421fa9-6125-4d7b-a489-4d18390fee45/original=true/117019443.jpeg" width="200"></td>
22
+ <td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/165003e2-cfa5-47a8-8392-dfecb9af6541/original=true/117214217.jpeg" width="200"></td>
23
+ <td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0b4852d6-8d88-41b2-a995-83ed0e65d4a1/original=true/117214225.jpeg" width="200"></td>
24
+ </tr>
25
+ <tr>
26
+ <td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd224564-c247-4397-8d1e-8193dd8c9e76/original=true/117214232.jpeg" width="200"></td>
27
+ <td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/b8bf819a-4d1e-4358-b569-986db888b641/original=true/117214475.jpeg" width="200"></td>
28
+ <td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/75ab46f9-e69d-4da3-b567-a8e82dd4b1cc/original=true/117214692.jpeg" width="200"></td>
29
+ </tr>
30
+ <tr>
31
+ <td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/753cc8d9-2280-4a11-b7a3-d1886a0c2d0f/original=true/117216354.jpeg" width="200"></td>
32
+ </tr>
33
+ </table>
34
+
35
+ ## Description
36
+
37
+ Klein 9B lcs consistency lora 20260415
38
+
39
+ RunningHub:
40
+
41
+ Lora:
42
+
43
+ <https://www.runninghub.ai/model/public/2044762492517228546?inviteCode=rh-v1279>
44
+
45
+ Online Workflow:
46
+
47
+ <https://www.runninghub.ai/post/2016109710943981570/?inviteCode=rh-v1279>
48
+
49
+ Klein 9B consist lcs preivew 20260328
50
+
51
+ preview version of lcs training method.
52
+
53
+ Further version is under development
54
+
55
+ Klein 4B Consistency Lora
56
+
57
+ This lora inspired by ByteDance [Heilos](https://github.com/PKU-YuanGroup/Helios) Easy Anti-Drifting method.
58
+
59
+ I implemented Frame-Aware Corrupt (Reference Latent Corrupt) latent and Color Statistics Loss.
60
+
61
+ In additionally, I use frequence split dataset which created by converting an image into frequence domain and split it into high freq image and low freq image.
62
+
63
+ Reference: low freq image, high freq image (0.8 dropout rate on high freq image)
64
+
65
+ Target: clean image
66
+
67
+ lr: 2e-4
68
+
69
+ scheduler: cosine
70
+
71
+ optimizer: adamw8bit
72
+
73
+ steps: 6000
74
+
75
+ Suggested Prompt:
76
+
77
+ {edit prompt}. Transform the image to realistic photograph. add realistic details to the corrupted image. restore high frequence details from the corrupted image.
78
+
79
+ Suggested Strength:
80
+
81
+ 0.5-0.7
82
+
83
+ Trained on 4B based. This lora aims to preserve image consistence across multiple input images. Control the lora via lora strength.
84
+
85
+ More strength means more consistence but it would also hurt edit abilities.
86
+
87
+ Different image might has different optimal lora strength.
88
+
89
+ Running Hub
90
+
91
+ Flux2 Klein 4B一致性lora对比 runninghub.ai/post/2032812180667633666/?inviteCode=rh-v1279
92
+
93
+ Flux2 Klein 4B一致性lora 不偏色 不偏移 runninghub.ai/post/2032817113190113281/?inviteCode=rh-v1279
94
+
95
+ Flux2 Klein 9B Consist Lora
96
+
97
+ HuggingFace:
98
+
99
+ <https://huggingface.co/lrzjason/Consistance_Edit_Lora>
100
+
101
+ Klein 9B Consis Lora
102
+
103
+ Suggested Prompt:
104
+
105
+ restore image details
106
+
107
+ Flux2 Klein 9B Consis Lora
108
+
109
+ Trained on 9B based. This lora aims to preserve image consistence across multiple input images. Control the lora via lora strength.
110
+
111
+ Suggested strength: 0.2~1.5
112
+
113
+ Use with any prompts.
114
+
115
+ More strength means more consistence but it would also hurt edit abilities.
116
+
117
+ Different image might has different optimal lora strength.
118
+
119
+ RH:
120
+
121
+ <https://www.runninghub.ai/post/2016109710943981570/?inviteCode=rh-v1279>
122
+
123
+ ======================QE2511================================
124
+
125
+ RH:
126
+
127
+ <https://www.runninghub.ai/model/public/2010085010870640641?inviteCode=rh-v1279>
128
+
129
+ Custom\_node:
130
+
131
+ <https://github.com/lrzjason/Comfyui-QwenEditUtils>
132
+
133
+ QE2511 Alpha is trained on QwenEdit 2511.
134
+
135
+ As same as previous lora, this lora helps to maintain image consistence.
136
+
137
+ The lora is trained using nf4 qlora.
138
+
139
+ V2 lora is trained for QwenEditPlus aka QwenImage Edit (2509)
140
+
141
+ It is recommanded to use with strength between 0.5 ~1.0 for multiple images
142
+
143
+ 0~0.5 for single image
144
+
145
+ It needs to use with workflow v4.0 [https://civitai.com/models/1939540](https://civitai.com/models/1939540?modelVersionId=2256750)
146
+
147
+ RH workflow:
148
+
149
+ QwenEdit Consistence Edit workflow v4.0
150
+
151
+ Experience link: <https://www.runninghub.ai/post/1965033238311915522>[/?inviteCode=rh-v1279](https://www.runninghub.ai/post/1965033238311915522/?inviteCode=rh-v1279)
152
+
153
+ QwenEdit Consistence Edit workflow v2.0
154
+
155
+ Solved pixels shift issue and keep qwenvl encode image ability.
156
+
157
+ RH workflow:
158
+
159
+ QwenEdit Consistence Edit workflow v2.0
160
+
161
+ Experience link: <https://www.runninghub.ai/post/1965033238311915522/?inviteCode=rh-v1279>
162
+
163
+ Trained with no ref image into qwenvl.
164
+
165
+ This lora typical to test the consistency with kontext like workflow.
166
+
167
+ It is suggested to use 0.45~0.5 lora stregth
168
+
169
+ Workflow:
170
+
171
+ <https://civitai.com/models/1939540/qwenedit-consistance-edit-workflow>
172
+
173
+ RH workflow:
174
+
175
+ It's extremely fun and I highly recommend it to you!
176
+
177
+ Workflow: QwenEdit Consistence Edit workflow
178
+
179
+ Experience link: <https://www.runninghub.ai/post/1965033238311915522/?inviteCode=rh-v1279>
180
+
181
+ ---
182
+
183
  Author: [xiaozhijason](https://civitai.com/user/xiaozhijason)
184
 
185
+ Model: [CivitAI Model Page](https://civitai.com/models/1939453?modelVersionId=2585269)
186
+
187
+ Archive: [CivArchive Page](https://civarchive.com/models/1939453?modelVersionId=2585269)
188
 
189
+ <!-- Version: 20260425_upload -->