
这是一个基于FLUX.1-fp8训练的,用于生成带网格的人物形象的Lora,推荐权重在0.8-0.9,
我使用的是Euler 20步,目前还在熟悉FLUX.1的跑法,所以可能不是最佳参数,
为了保证出图流程,我的范例图都叠加使用了使用了TensorArt上的FLUXREALnime,你也可以尝试其他的叠加效果,[注意,REALNIME是FLUXREALnime的触发词,不是本模型的,如果你想参考本模型示例图的提示词格式,且不叠加FLUXREALnime,请记得删掉它]
https://tensor.art/models/773433368722432100/FLUXREALnime-V1.0
叠加情况下基本是FLUXREALnime 0.6-0.7 NGYOU 0.8-0.9
有兴趣的大伙可以下载玩一下
This is a LoRA model, which is trained with FLUX.1-fp8 as the base model, that could generate a gridded body image. The recommended weight range is 0.8-0.9.
I’m using Euler with 20 steps. Currently, I’m still getting familiar with the FLUX.1 setup, so these might not be the optimal parameters yet.
To ensure a stable image generation process, all my sample images have been layered with TensorArt’s FLUXREALnime. [Note: "REALNIME" is the trigger word for FLUXREALnime, not for this model. If you want to use the prompt format from the sample images without layering FLUXREALnime, please remember to remove "REALNIME."]
https://tensor.art/models/773433368722432100/FLUXREALnime-V1.0
For layering, I typically set FLUXREALnime to 0.6-0.7 and NGYOU to 0.8-0.9. Feel free to try other LoRA layering with it.
Have fun with it.
描述:
训练词语: NGYOU
名称: NGYOU-EPOCH17.safetensors
大小 (KB): 65708
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success