SevrMix – V2
| Run Model | 使用GRAVITI Diffus 关于GRAVITI Diffus |
| Model Name | SevrMix |
| Model Type | Checkpoint |
| Base Model | SD 1.5 |
| Version Name | V2 |
| File Name | sevrmix_v2.safetensors |
| SHA256 | 2b0cf24102 |
| NSFW | False |
| Trigger Words | |
| Tags | character, photorealistic, woman, girls |
| Metadata | fp = fp32 size = full format = SafeTensor |
| Size | 7.34 GB |
| Version Description | None |
| Source | https://civitai.com/models/36153 |
| Author | Showevr |
这是我自己经常使用的混合模型，创建这个模型的动机是：我需要一个可以绘制中国古建筑的大场景模型。
如果画面出现人脸扭曲，这是因为因为这个模型偏向于生成大场景，人物在图像中的占比很小，没有足够的画面渲染人脸。请你用Hires. fix放大图像，另外，我建议你在选择放大算法的适合，选用R-ESRGAN 4x+，并将Denoising strength设置为0.5以下，我经常使用0.3~0.5左右的值。
如果你想实现中国少数民族-苗族的服装效果，你需要额外下载下面的Lora：
苗族服装 | Hmong costume | Stable Diffusion LORA | Civitai
——————————————————————————————
This is a hybrid model that I frequently use, and the motivation behind creating this model is that I needed a model that could generate large-scale scenes with Chinese ancient architecture.
If the faces in the image appear distorted, it’s because this model tends to focus on generating large scenes, and the proportion of characters in the image is relatively small, resulting in insufficient rendering of the faces. I recommend using Hires. fix to enlarge the image, and for the upscaling algorithm, I suggest using R-ESRGAN 4x+ with a Denoising strength of 0.5 or less. I often use values around 0.3 to 0.5.
If you want to achieve the effect of the costume of a Chinese ethnic minority – the Miao people, you need to download Lora separately:
相关文章
