This LoRA changes the style of the generated image of Nabylon-v1.2 or other anime models. Applying this LoRA to an anime model can bring the representation of a person closer to a realistic look, such as accurate shading and realistic cloth representation.
It has been trained combination with Nabylon-v1.2 https://civitai.com/models/26444
Apply this LoRA with a small weight; applying it with weight around 1.0 will provide images with too much effects. Note that this LoRA is not designed to generate fully realistic style images, so if you are aiming to generate realistic images, it is recommended to use other models such as Chilloutmix.
Weight guideline
-
0.2 Sufficient and improves the representation of shadows and costumes.
-
0.4 Starts to affect the shape of the face. In particular, the nose will be drawn with a more realistic shape.
-
0.6 Sometimes the representation is between anime and realistic-looking, which is not common.
-
0.8 Too large. The expression becomes quite realistic (but not completely realistic).
描述:
Snapshot of the model under training.
train image set ver.3 (35200 images), large model
训练词语:
名称: NabylonRS-3L-epoch018.safetensors
大小 (KB): 197145
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
