[LoRA] line illustrationa
模型简介：这是做复古线条彩色漫画风格的lora（有点点像《丁丁历险记》漫画的风格，但是颜色层次和细节又相对比较丰富，我也不知道这种风格怎么称呼比较合适），训练图用了73张训练图,
底模用的是pasteldiffusedmix_v20.safetensors，尺寸是512x768，
画面有人物的时候，最好勾上插件After Detailer,出的图尺寸不宜太大，会有错位；
试了其他二次元的其他大模型，一般度能适用；权重看着调就好；
The following is machine translated
This is to do retro line color comic style lora (a little bit like The Adventures of Tintin comic style, but the color level and detail is richer, I do not know how to call this style is more appropriate), the training map used 73 training pictures,
The base model is pasteldiffusedmix_v20.safetensors, size is 512x768.
When there are characters on the screen, it is best to hook up the plug-in After Detailer, the size of the figure should not be too large, there will be misalignment;
Tried other secondary other large model, the general degree can be applied; weights look at the adjustment is good;
~~~~~~~~~~~~~~~~~~~~~~~o(*￣︶￣*)o~~~~~~~~~~~~~~~~~~~~~~~
示例提示词和参数：
Prompt: multiple boys, tree, 2boys, shirt, backpack, black hair, male focus, shorts, outdoors, white shirt, bag, pointing, short hair, short sleeves, smile, black shorts, day, vest, standing, socks, closed eyes, open mouth<lora:line illustrationa_20230901192549:0.83>
Negative prompt: EasyNegative
Steps: 20, CFG scale: 7, Sampler: DPM++ 2M SDE Karras, Seed: 388400863, Size: 512x800, Model: pasteldiffusedmix_v20, Version: v1.5.2, Model hash: 6dcdfda71d, "EasyNegative: c74b4e810b03", Hires upscale: 2, ADetailer conf: 30, Hires upscaler: Latent, ADetailer model: face_yolov8n.pt, ADetailer version: 23.5.19, Denoising strength: 0.7, ADetailer mask blur: 4, ADetailer dilate/erode: 32, ADetailer inpaint full: True, ADetailer inpaint padding: 0, ADetailer denoising strength: 0.4, "line illustrationa_20230901192549: 0a576e614ff7", Clip skip: 2
页:
[1]