Sentinel7 commited on
Commit
2b7e8f0
·
verified ·
1 Parent(s): c8e6658

Upload 1954733/2450945/README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. 1954733/2450945/README.md +66 -6
1954733/2450945/README.md CHANGED
@@ -1,25 +1,85 @@
1
  ---
2
  license: other
3
  tags:
 
 
 
 
 
 
4
  ---
5
 
6
- # Unknown - 2450945
7
 
8
- **Model Type**: Unknown
9
 
10
- **Base Model**: Unknown
11
 
12
  **Trigger Words**: None
13
 
14
- **Tags**: None
15
 
16
  ## Description
17
 
18
- No description available.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19
 
20
  ---
21
 
22
- Author: [Unknown](https://civitai.com/user/Unknown)
23
 
24
  Model: [CivitAI Model Page](https://civitai.com/models/1954733?modelVersionId=2450945)
25
 
 
1
  ---
2
  license: other
3
  tags:
4
+ - anal
5
+ - butt
6
+ - concept
7
+ - nsfw
8
+ - penis
9
+ - woman
10
  ---
11
 
12
+ # Asshole - Wan 2.2 T2V/I2V 14B-Low I2V-14B - 2450945
13
 
14
+ **Model Type**: LORA
15
 
16
+ **Base Model**: Wan Video 2.2 I2V-A14B
17
 
18
  **Trigger Words**: None
19
 
20
+ **Tags**: anal, butt, concept, nsfw, penis, woman
21
 
22
  ## Description
23
 
24
+ I2V
25
+ ===
26
+
27
+ I2V has finally landed, performing much better than using the T2V lora with the I2V model. Trained on the exact same dataset, excluding images as they cannot be used to train i2v. The low noise model has been trained at the native resolution of the dataset for maximum fidelity (1024x1024).
28
+
29
+ The concepts and prompts are the same as the T2V version.
30
+
31
+ T2V
32
+ ===
33
+
34
+ Trained on a quite large collection and a lot of different concepts, the main one being women lying on their back showing their asshole. Regularization data were also mixed in to prevent bias.
35
+
36
+ The model seems to perform well in a wide variety of scenes and has the potential to perform well in much more tasks than the one it was intended for.
37
+
38
+ ### Training/Dataset parameters
39
+
40
+ * 259 \* 512x512 videos of 3 seconds.
41
+ * 189 \* 1024x1024 images.
42
+ * Rank 128.
43
+ * 18 epochs on the high noise.
44
+ * 22 epochs on the low noise.
45
+ * Higher repeats on concepts with less data.
46
+ * Learning rates started at 0.0001 for both, and the last ~20% of training ran at about 0.00002.
47
+ * Videos were sliced and cropped using a custom tool I wrote.
48
+ * Captioning:
49
+
50
+ + Qwen2.5-VL was used for the initial captioning.
51
+ + The system prompt was adapted for each different concept to provide the model with a context and instructions on what to look for.
52
+ + Manually reviewed with minor corrections.
53
+ * Trained using diffusion-pipe on a single 5090.
54
+
55
+ ### What it what trained on:
56
+
57
+ * Women in piledriver position
58
+
59
+ + Spreading her buttocks with own hands
60
+ + Viewer spreading her buttocks
61
+ * Close-up of vaginas from under
62
+ * Penis anal insertion
63
+ * Penis anal pull-out
64
+ * Penis anal thrusting
65
+ * Side views
66
+ * POV views
67
+ * A variety of pubic hairiness, but I am somehow still unable to control it
68
+ * Ejaculation, but hard to control
69
+ * Faces were mostly cut out of frame to minimize character bleeding
70
+
71
+ ### What it cannot do:
72
+
73
+ * Do NOT expect good results on standing poses. There is none in the dataset.
74
+
75
+ See the prompts associated with the samples.
76
+
77
+ **Prompt geniuses please share your work.**
78
+ -------------------------------------------
79
 
80
  ---
81
 
82
+ Author: [cnovice](https://civitai.com/user/cnovice)
83
 
84
  Model: [CivitAI Model Page](https://civitai.com/models/1954733?modelVersionId=2450945)
85