AnimeXL-xuebiMIX的最新版本，相比之前版本大幅度地改善了图片的光影，让图片层次感和氛围感更强
另外本模型已经融合了LCM-LORA，无需再额外调用，因此使用本模型时推荐以下设置：
1.采样算法选择Euler-a（根据琥珀大佬的意见，实测比LCM采样效果好很多）
2.迭代步数6-10即可产出高质量图片，大大加快出图速度
3.高分修复算法可自行选择尝试，迭代步数8-10为佳，重绘幅度根据算法不同自行尝试
4.CFG Scale设置3-4为佳
当然，这模型也并非完美，目前主要情况是肢体细节仍然不太理想，正在研究改进，希望大家能多多测试返图，谢谢
Description
FAQ
Comments (9)
it is possible to use it with rtx 3060 12gb, CPU ryzen 7, 16Gb ram?
Any sdxl model should work. It runs (with --medvram) even on my rtx 3060 6gb, same cpu and ram
Yep, as xlevovix mentioned, using the --medvram flag when starting a1111 may work. Comfy UI also works for me without any additional configurations - on an 8gb card.
The RAM might be an issue. Like Ocean3 says, use medvram and might work fine.
I'm using 10GB on AMD, still uses almost all my 32GB RAM. It gonna depend how much complexity you want to add. Using the refiner takes a lot of RAM.
I use no VRAM arguments, and this checkpoint at 1024x1024 uses 6.6GB total of my 10GB on a 3080. I have 32GB RAM and it uses 0% of that with ComfyUI, except for what is necessary for the browser to run. I am getting little white speckles all over the character though so I could be having a prompting issue with that.
@Ocean3 that's actually because ComfyUI default setting is medvram. You can add a --highvram argument to use more of it if you want.
@xlevovix795 I wouldn't use any of the unusually large ones, (bigger than 6.some GB)
@xlevovix795 for sdxl should use --medvram-sdxl
List of merged models?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.