In the v4.0 samples,the pictures on the left and right are generated based on v4 and v1, respectively.
v4.0的samples中,左边和右边的图分别是根据v4与v1生成的
compared with v1.0,v4.0 keeps more features of the style and the lines at the junction of light and dark is clearer.And v4.0 need lower weight( just about 1)
与v1.0相比,v4.0能保留更多的画风的特征,明暗交界处的线条也更明显.v4.0要求的weight也更低(1就差不多了).
And in v3.0,I add new source material in the train set ,so it may be anothers paint style with some difference to v1.0 and v4.0
Description
训练集主要是焦茶的插画和其他人模仿焦茶画风的作品,焦茶 - pixiv
The training set is mainly about the illustrations of Cogecha and other people's works imitating Cogecha's painting style
我推荐的权重是1.1到1.25
The lora weight I recommend is 1.1 to 1.25
FAQ
Comments (18)
Some prompts or trigger words ?
Tried it and looks like no specific trigger words are needed, just including this LoRA changes the style of the image.
no trigger words,all things you have to do is just load this LoAR
I've been playing with this a bit and it seems if a particular image isn't picking up the ligne claire, bumping the strength up to 1.1-1.2 seems to help. This will overdo it for other images though so I put it back down after. Some other prompts can help too, see the description and example images of this Ligne Claire model: Ligne Claire Anime | Stable Diffusion Checkpoint | Civitai). Also, booru tags ligne claire and the artist cogecha do exist but don't seem to affect it much, but I guess are worth noting as possible tags to try.
I think this one is fake. Discord suggested a few of them got some fake images on top of renamed models.
@decten76 I have tried this model and it seems to work, whether or not OP is the author I can't say, though his other LoRA does have images with metadata so I'm not sure I'd jump to thinking this is fake.
@fras If your lora cannot be used, you can try to update your lora plug-in or webui to the latest version. This is not a fake lora!!!,😓
@decten76 If your lora cannot be used, you can try to update your lora plug-in or webui to the latest version. This is not a fake lora!!!,😓
OP Post prompts:
1girl,white background, cowboy shot,normal+shorts+navel+white shirt
Negative prompt: ((coat,cape)),sketch, duplicate, ugly, huge eyes, text, logo, monochrome, worst face, (bad and mutated hands:1.3), (worst quality:2.0), (low quality:2.0), (blurry:2.0), horror, geometry, bad_prompt, (bad hands), (missing fingers), multiple limbs, bad anatomy, (interlocked fingers:1.2), Ugly Fingers, (extra digit and hands and fingers and legs and arms:1.4), crown braid, ((2girl)), (deformed fingers:1.2), (long fingers:1.2),succubus wings,horn,succubus horn,succubus hairstyle, (bad-artist-anime), bad-artist, bad hand,
Steps: 22, Sampler: DPM++ 2M Karras, CFG scale: 8.5, Seed: 831741454, Size: 448x704, Model hash: e4b17ce185, Denoising strength: 0.5, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires steps: 20, Hires upscaler: Latent (antialiased), AddNet Enabled: True, AddNet Module 1: LoRA, AddNet Model 1: AK_狮蝎3(normal+shorts+navel+white shirt_dirge+dress+pantyhose_sargon+bodystocking_sargon+bodystocking+face veil)(e23c8bf99159), AddNet Weight A 1: 1, AddNet Weight B 1: 1, AddNet Module 2: LoRA, AddNet Model 2: onelastlastcogecha(8f8a48dd3c60), AddNet Weight A 2: 1, AddNet Weight B 2: 1
unlikely working. Images presented without prompt and setting.
If your lora cannot be used, you can try to update your lora plug-in or webui to the latest version. This is not a fake lora!!!,😓
it look very good, can you tell me what is your model?
When future generations appreciate the media-published artworks of such style, and inquire as to the hand that wrought them, the elder will tell them of him, of the traces he left in the world, and how he shall forever be remembered.
Wish you a pleasant journey in heaven.
当后人欣赏着媒体发表的这种风格的艺术品，询问出自谁之手时，老一辈们会告诉他们他，他在世界上留下的痕迹，以及他将如何被永远铭记。
愿在天堂的旅途愉快。
这个模型训练的太好了,它不怎么会改变人物形象但却能把整体绘画手法都改了一遍.
另外我想请问一下训练lora模型主要用什么参数才能像这个模型一样着重改变画风而不是人物?
我自己训练出来的全都在变成其他人
I want to learn, too
me too
moxin helltaker burtonesqueStyle
请问有讨论模型融合和lora相关的群吗？我想参与技术讨论。
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.