v3感觉练过头了,使用建议搭配质量词or画师串,与其它大模型生图情况不一,效果不好的话降权重。
发现一画师串(sketch lines,mixed media,conceptual art, z3zz4, dino \(dinoartforame\), ichika \(ichika87\), toosaka asagi, nekojira, rhasta, rella)效果超棒,欢迎大家使用。
发现一画师串((bm tol:1.3),(ask \(askzy\):1.1),applepie \(12711019\),mr. j.w,)效果很强,生成的图自带光影平涂非常nice不加Lora效果也在,那种平涂效果就是此Lora追求的最终效果。
质量词推荐:(shadow,masterpiece,best quality,good quality,very aesthetic,A shot with tension,impactful picture,best quality,very aesthetic,extremely delicate and beautiful,huge filesize,amazing quality,solo_focus,sunlight,reflective,masterful details,intricate details,realistic,(bm tol:1.3),(ask \(askzy\):1.1),applepie \(12711019\),mr. j.w,close-up,depth of field,Focus on the face,ray tracing,lens flare,aesthetic,looking at viewer)
翻译:(一个有张力的镜头,有冲击力的画面,最好的质量,非常美学,非常精致美丽,巨大的文件大小,惊人的质量,solo_focus,阳光,反射,精通细节,复杂的细节,逼真,(bm tol:1.3),(问\(askzy\):1.1),applepie \(12711019\),mr。近景,景深,脸部对焦,光线追踪,镜头光晕,美学,看着观众)ps:有镜头控制词(Focus on the face,looking at viewer)
与词冲突的话记得检查删除哟。
Version Detail
基础模型 XL
6
lora基于天光征兆进行练行,在网上找了43张平涂的图,高清放大至2K后训练而出,直出效果还行,建议搭配质量词(没有生成信息的图片为Lora测试生成的图),lora权重推荐0.8(待生图补充),具体等我用这lora搭配各位大佬的底膜试一试.具体效果生图后
Project Permissions
Use in 吐司 Online
As a online training base model on 吐司
Use without crediting me
Share merges of this model
Use different permissions on merges
Use Permissions
Sell generated contents
Use on generation services
Sell this model or merges










