flux2-klein-9b-clothes-slider-v4.safetensors

40 MB · SafeTensor · bf16

Lidmaatschap vereist

<clothes-slider4-flux2>

Version 4 (NEW) Weight ranges from -2 (heavy clothing) to 2 (no clothing). 权重为-2（厚重衣物）~2（无衣物） This LORA only enhances the model's existing capabilities, so it is recommended to use it in combination with other LORAs. 这个lora只是强化模型原有能力，所以建议搭配其它lora使用 The prompt in the example image is (for reference only): replace all of my clothes while keep maintaining the original character and pose unchanged 示例图片中的提示词为（仅供参考）：替换我所有的服装，同时保持原有角色特征和姿势不变 Version 3 I have re-implemented the loss function and added more prompts to mitigate interference. I hope it works~ 我重新实现了损失函数，并添加了更多的提示词来减弱干扰，但愿有用吧~ Version 2 I optimized the loss function composition, which is now more stable than before and can generate NSFW content. However, since this slider was trained using text without introducing new knowledge, it struggles to generate physiological structures effectively. It can be used in combination with other NSFW LORA models. 我优化了损失函数的构成，现在比之前更稳定，并且可以生成NSFW的内容，但是由于这是通过text训练的slider，并没有额外引入新的知识，所以无法很好生成生理结构，可以搭配其它nsfw的lora使用 This LoRA gives you a smooth outfit-coverage slider for characters. Use negative weights for more covered / warmer looks (long sleeves, long pants, heavier layers), and positive weights for more exposed / lighter looks (short sleeves, shorts, swimwear-style outfits). The change is designed to feel gradual across the scale rather than like an on/off switch. It supports weights from -10 to +10, but very large values may cause odd or exaggerated results—dial back if things start to look weird. 这个 LoRA 提供一个顺滑的服装覆盖度滑条。权重为负值时更偏向穿得更厚、更包裹（长袖长裤、外套感更强）；权重为正值时更偏向穿得更清凉、更露（短袖短裤、泳装风格）。整体变化是渐进的，更像连续调节而不是突然切换。 权重支持 -10 到 +10，但绝对值太大可能会出现夸张或奇怪的效果，遇到这种情况把权重往回调即可。

SHA256

D9AFF59A06F9E68F883121F72C6060520061AEE3F84AC15A051586C7A6EC87A9

ModelTrends.ai model-ID

#24432