Preface: I have always liked the texture of wool felt style, especially the more realistic felt style. However, most of the styles on the market are more cute, which differs a lot from my expectation of realism. So I had the idea to train this Lora.
Before this, I trained the felt Lora of version SD1.5, but the effect was not very stable; this time I reorganized and optimized the training materials, achieved SDXL 1.0 version, with enhanced generalization, capable of generating images of themes such as characters, animals, plants, etc., supporting natural language descriptions. Highly recommended for wool felt enthusiasts~
Recommended keywords: Support natural language or phrases, realistic style preferred, for those who like cute styles can add chibi (refer to examples).
Trigger words: woolfeltkk, MN
Recommended parameters:
LORA weight: 0.8-1
Large models: Dream Industrial XL, sd_xl_base_1.0 and other realistic models, can also be used with other Loras.
Sampler: DPM++ series
Iteration steps: 25-35
CFG: 7
CLIP: 1
Resolution: Above 1024*1024
Welcome back for more images^^
前言：我一直很喜欢羊毛毡质感的风格，尤其是偏真实感的毛毡风格。而现在市面上大多数偏可爱的，跟我预期写实感差很多，所以我有了训练这款lora的想法。
在此之前我训练过SD1.5版本的毛毡lora，但效果不太稳定；这次重新梳理和优化训练素材，实现了SDXL1.0版本，泛化性有加强，可以生成人物、动物、植物等题材的图片，支持自然语言描述。非常推荐毛毡爱好者使用~
推荐提示词：支持自然语言或词组，写实为主，喜欢可爱风格可以加chibi（可以参考示例）
触发词：woolfeltkk,MN
推荐参数：
LORA权重：0.8-1
大模型：筑梦工业XL、sd_xl_base_1.0 等写实模型，也可搭配其他 Lora 使用。
采样器：DPM++系列
迭代步数：25~35
CFG：7
CLIP：1
分辨率：1024*1024以上
欢迎大家返图^^