Sampler：DPM++2M Karras     steps： 38     512X1024   cfg：7Hires. fix： X2，SwinIR 4x step：25 Denoising：0.26hope u like it! more my model  <a target="_blank" rel="ugc" href="https://www.liblib.art/userpage/1e3c4c0fc33545f89056e8b46523d3da/publish/model">https://www.liblib.art/userpage/1e3c4c0fc33545f89056e8b46523d3da/publish</a><a target="_blank" rel="ugc" href="https://www.liblib.art/userpage/1e3c4c0fc33545f89056e8b46523d3da/publish￼￼more"> </a>more my pics <a target="_blank" rel="ugc" href="https://www.liblib.art/userpage/1e3c4c0fc33545f89056e8b46523d3da/publish/image">https://www.liblib.art/userpage/1e3c4c0fc33545f89056e8b46523d3da/publish/image</a>If you are unable to replicate my example image, the reason is that the random seed of the sd-webui where I generated the image is generated from the CPU, and the default setting is GPU. You can try changing it.

繁花|写真-10

10_v10.safetensors

v1.0

Flowers | Photo Album - 10

Another inventory, compared to the elaborate09, the color is lighter and there is more play of light and shadow. The style is similar to the earliest large model of mine <a href="https://www.liblib.art/modelinfo/3521e968cbcb4b76afbca28d7221ae8d" target="_blank">vigee-asian_realistic</a> 1.2.X version. However, the rendering is completely different, also using completely different blend materials.There's a slight color change, reduced a bit of yellow, making the skin look a bit whiter. Correspondingly, the overall tone leans slightly towards blue and is colder.As for Lora's compatibility, I really haven't tested it because I hardly use Lora. Even if I change the face shape of the locally generated image, I call i* ***ng the ADetailer plugin. If there is any test, I hope to receive some feedback.

Files with SHA256: c3b39b9bb647c8ab70e154037fd2933ed0a9bf1c835eeacef62b68f661b52cf0