1、第一我用了在线平台
https://www.liblib.art/sd?info=oneclick&id=1716514193089
2、模型上我用了两个模型
2.1 欧式单体建筑模型
https://www.liblib.art/modelinfo/7f60726bdb2c41c1ac009b2a678945e2
Conceptual art
European architectural design
top-down perspective
正向触发词：
Arab palace,Conceptual art,European architectural design,top-down perspective,rock,gradient,gradient_background,grass,stairs,scenery,sitting,grey_background,water,1girl,flower,weapon,status,tree,brown_hair,outdoors,pillar,standing,waterfall,ruins,
负面触发词：
FastNegativeV2,(bad-artist:1),(worst quality, low quality:1.4),(bad_prompt_version2:0.8),bad-hands-5,lowres,bad anatomy,bad hands,((text)),(watermark),error,missing fingers,extra digit,fewers digits,cropped,worst quality,low quality,
然后它的checkpoint实际上是
这个
它本身实际上是一个lora
2.2然后我调用了第二个lora
https://www.liblib.art/modelinfo/b54aca58ee3f447987f5ddfc7dfe84f1
触发词是这个
pixel art
pixel
2.3 最终触发词
Arab palace, Conceptual art, European architectural design, top-down perspective, rock, gradient, gradient_background, grass, stairs, scenery, sitting, grey_background, water, 1girl, flower, weapon, status, tree, brown_hair, outdoors, pillar, standing, waterfall, ruins,pixel,pixel art,pixel,Pixel style,
最终负向词：
FastNegativeV2,(bad-artist:1),(worst quality, low quality:1.4),(bad_prompt_version2:0.8),bad-hands-5,lowres,bad anatomy,bad hands,((text)),(watermark),error,missing fingers,extra digit,fewers digits,cropped,worst quality,low quality,
2.4 采样参数
2.5 Layer diffusion参数
2.6 最终全参数记录
Arab palace,Conceptual art,European architectural design,top-down perspective,rock,gradient,gradient_background,grass,stairs,scenery,sitting,grey_background,water,1girl,flower,weapon,status,tree,brown_hair,outdoors,pillar,standing,waterfall,ruins,pixel,pixel art,pixel,Pixel style, Negative prompt: FastNegativeV2,(bad-artist:1),(worst quality, low quality:1.4),(bad_prompt_version2:0.8),bad-hands-5,lowres,bad anatomy,bad hands,((text)),(watermark),error,missing fingers,extra digit,fewers digits,cropped,worst quality,low quality,,nsfw Steps: 24, Sampler: DPM++ 2S a, CFG scale: 7.0, Seed: 1896815881, Size: 512x512, Model hash: 7c819b6d13, Model: majicMIX realistic 麦橘写实_v7.safetensors, VAE: automatic, Clip skip: 2, RNG: CPU, Lora 1: Pixel_像素世界, Lora Hash 1: b903e2f916, Lora Weight 1: 0.8, Lora 2: 单体建筑V2(欧式）, Lora Hash 2: e722cfa3ce, Lora Weight 2: 0.8, TI hashes: "FastNegativeV2: a7465e7cc2, bad-hands-5: aa7651be15", layerdiffusion_enabled: True, layerdiffusion_method: FG_ONLY_ATTN_SD15, layerdiffusion_weight: 1.0, layerdiffusion_ending_step: 1.0, layerdiffusion_fg_image: False, layerdiffusion_bg_image: False, layerdiffusion_blend_image: False, layerdiffusion_resize_mode: Crop and Resize, layerdiffusion_fg_additional_prompt: , layerdiffusion_bg_additional_prompt: , layerdiffusion_blend_additional_prompt: , Version: f0.0.17v1.8.0rc-latest-280-gaaf85f19
2.8 最终结果
3. 如何导入RPG Maker
我最后实际上是没用使用Map Tile的方式，最后等于使用了事件触发的图库位置
我直接等于是把图像扔到：E:\hgames\playWithMe\测试游戏\img\characters
目录里面去了，然后这个目录有一个优点是，用!$作为开头的图像，会有一个约定俗称，我测试了一下，PNG的透明图像，横向是3个，纵向是4个，一共可以存储12个位元的图像
最后就是这样一张东西
紧接着进入RMMZ
事件触发模式
选择图像
然后就可以了
编写触发时间，进入房间即可
oblique的top-down视角，这个记住了，可以作为关键字
最后是运行效果
图片的阴影啊这些还需要加工啥的，这些需要交给PS啥的
但大的东西是可以省很多事儿了
感谢LiblibAI平台，确实省事儿，之后可以在本地搞一个
朋友提醒还有一种叫做：
isometric投影会比oblique舒服些
3、最后是视频效果：