English/中文

Are you still bothered by the expensive films of Polaroid? Now, with this LoRA, you can easily have photos with the style of Polaroid Onestep2 without spending a lot of money on films. Whether it's landscapes, portraits, or other subjects, you can easily create unique Polaroid Onestep2 effects.

This LoRA is trained on photos taken with Polaroid Onestep2 and can reproduce its unique style quite well.

Instructions:

• Trigger word: chadpo

• Weight: 0.8-1

• Recommended positive words: masterpiece, highres, best quality, highly detailed, original, delicate background, extremely detailed 8K wallpaper

• Interesting trigger word: gradient sky

• Recommended negative words: border

• Recommended models: cyberrealistic_v31, Realistic Vision V2.0 (this LoRA is trained on SD1.5, so it can be used with most large models based on SD1.5. Different models may produce slightly different results, but overall they are acceptable. Refer to example and test images. Some models may require lowering the weight)

• Example settings: Steps: 20, Sampler: DPM++ 2M SDE Karras, CFG scale: 7, Denoising strength: 0.5, Clip skip: 2, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ (training set images are 1024x1024, so Hires.fix is not necessary to generate good images)

Creating this LoRA has taken a lot of effort, and I hope to receive more comments and feedback, as well as images created using this model. If you have any questions or suggestions regarding this LoRA, please feel free to contact me. I look forward to your feedback!

还在为拍立得昂贵的胶片而烦恼吗？现在，有了这个LoRA，你可以轻松地拥有拍立得Onestep2风格的照片，而无需花费大量金钱购买胶片。无论是风景、人像还是其他主题，你都可以轻松创造出独特的拍立得Onestep2效果。这个LoRA制作花费了我很多精力，希望大家能点个爱心、返图分享一下你的作品。

该LoRA以Polaroid-Onestep2拍摄照片为训练集，可以较好的还原其独特风格。

使用方法：

• 触发词：chadpo

• 权重：0.8-1

• 推荐正面词：masterpiece, highres, best quality, highly detailed, original,Delicate background,extremely detailed 8K wallpaper,

• 有意思的触发词： gradient sky

• 建议负面词：border,（负面词请自行尝试，部分例图未使用任何负面词，负面词可能会改变风格）

• 推荐模型：cyberrealistic_v31、Realistic Vision V2.0（该LoRA以SD1.5为底模训练，故可适用于大多数以SD1.5为基础的大模型，不同模型产生的效果会有少许差异，总体来说都还行，具体可见例图、测试图，部分模型可能需调低权重）

• 例图设置：Steps: 20, Sampler: DPM++ 2M SDE Karras, CFG scale: 7, Seed: 2576664679, Denoising strength: 0.5, Clip skip: 2, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+（训练集图片大小为1024x1024，故不用高清修复也可以生成不错的图像）

• 注意：边缘质感控制不良(可用border作为负面词抑制)，部分景物无法生成，有时会出现噪点，建议多roll(属于是模拟了拍摄失败时的情景了；若想生成大尺寸图像，建议先不开启高清修复，先快速生成一批图片，再挑选记下构图合适的种子，最后开启高清修复重新生成相应种子图像)

如果你对该LoRA有任何疑问或者建议，随时都可以联系我，期待你的反馈！