Type | |
Stats | 129 185 106 |
Reviews | (21) |
Published | Feb 10, 2025 |
Base Model | |
Training | Steps: 1,335 Epochs: 10 |
Usage Tips | Clip Skip: 2 Strength: 0.8 |
Hash | AutoV2 7335C37420 |
Well, here goes my another attempt at LoRA. This LoRA is based on style of the mangaka Tanaka Yutaka. He is known for his ecchi and hentai works.
I trained this LoRA on different black & white panels from his manga "Itoshi no Kana" (80 pictures). Some panels were edited to make training easier for neural network.
That girl with light purple hair is a ghost, BTW.
Attention!
0) Preferable weights 0.6 - 1.0 (depends on checkpoint and other LORAs used in generation)
1) No additional tags needed for using this LoRA. Just add and it's ready to go.
2) It displays artist's style even with score and quality tags.
3) As my testing shows, this LoRA works with every checkpoint. If you want to use this LoRA on other checkpoints, then you should adjust the weight of this LoRA (0.8 is good enough for most cases).
4) Yes, you can use it to generate hentai images on Civitai.
5) This mangaka likes, no, he LOVES using close-ups of faces. So if you have issues, then add "close-up" to negatives. Also he dislikes drawing detailed backgrounds.