使用中文提示词+60张厚涂特写风格的图练出的wan的lora，用的是自然语言，使用时在提示词的开头写上“数字厚涂风格的特写镜头”，然后跟上你的画面描述即可，推荐使用中文提示词尝试效果。使用的时候，高低噪模型都要使用，我出的示例图都是：10步高噪声+10步低噪声，shift 3生成的，权重推荐是0.8-1，示例图中有不同权重的效果对比<hr />The LORA model trained using Chinese prompts and 60 thick-painting close-up style images. Utilise natural language prompts by prefacing your description with ‘数字厚涂风格的特写镜头’ followed by your scene description. We recommend experimenting with Chinese prompts for optimal results.When using the model, employ both high-noise and low-noise settings. My sample images were generated with: 10 steps high-noise + 10 steps low-noise, using Shift 3. Recommended weight values range from 0.8 to 1. The sample images include comparisons of effects at different weight settings.

Files with SHA256: 866d097915005fcb399273c051799475b009cba64e2b474afe8e1d46068eb850