For V2
prompts: score_9,score_8_up,score_7_up,
Neg prompts: score_6,score_5,score_4,artist name,signature,watermark,ugly,worst quality,poor details,blurry,bad anatomy,bad hands,error,extra limb,missing limbs,abs, simple background, dark background, white background,
Description
FAQ
Comments (9)
v1.5 - Every picture has numerous white dots in the generation and looks washed out compared to 1.0. Am I missing something?
may be missing select VAE
@ppkkmoon Any idea on why the white dots? I did have a vae selected rather than just automatic, see screenshot - https://civitai.com/posts/2021380
@Sharinel may be try c9 vae, https://civitai.com/models/152040?modelVersionId=173672
大佬，你好，我用你的V10模型进行了微调，不知道是否可以在C站发布呢？Hello, big guy, I have made fine adjustments with your V10 model, I wonder if I can publish it in C station?——Youdao
no luck on any of these models even with different vae results are grey washed out and look like a depth map in comfyAi
<lora:realistic characters_0001: > What is it? Where to download? ......
[New review]
Not sure what issues I had before, but this is often my go-to now before PonyDiff, it can give cleaner results at times, and plays nicely with a lot of LoRAs.
[OLD REVIEW]
I get nothing but trash results from this, even copying the OP results. Human bodies end up elongated to a ridiculous degree, have numerous extra limbs, and overall look entirely nightmare fuel at any combination of CFG and steps, with any sampling method.
Avoid this checkpoint entirely.