推薦採樣器（sampler）： Euler a / SGM_uniform (推薦 recommend)
步數（steps）：8-10
這是個寫實美女網紅模型與動漫模型融合後,再融入"細節LORA"的全新模型。會想製作這個模型的起因是本人偏好寫實模型(所以融合比例偏高),但同時又想擁有3D CG般的皮膚，於是就嘗試將這寫實模型融合高畫質動漫CG模型，然後再次融合"細節LORA",所以整體畫質與3D CG皮膚表現以寫實模型來說堪稱完美(個人觀感,請勿批評,畢竟每個人審美觀不同)。所以本模型既能畫寫實美女、風景,同時也能畫動漫人物,而且細節也不差；只需參考我的帖子中有關增強"動漫人物環境設定提示詞",便可達到相當不錯的畫質感。而這次訓練的人物針對皮膚部分加強處理,看上去會變得光滑細膩質感。
This is a brand new model that merge a realistic female internet celebrity model with an anime model, and then merge "detailed LOARA" elements. The reason I wanted to create this model is that I prefer realistic models (hence the high merge ratio), but at the same time, I wanted 3D CG-like skin. So, I tried merge this realistic model with a high-quality anime CG model, and then further merge it with "detailed LOARA." Therefore, the overall image quality and 3D CG skin rendering are practically perfect for a realistic model (this is just my personal opinion, please don't criticize, as everyone's aesthetic sense is different). This model can depict realistic beauties and landscapes, as well as anime characters, and the details are quite good; just refer to my post on enhancing "anime character environment setting prompts" to achieve a very good image quality. The character trained this time has enhanced processing on the skin area, making it look smoother and more delicate in texture.
另外基礎的寫實模型自帶"NSFW"功能,加上融合比例較重,所以依舊保留了"NSFW"功能,對於喜歡"NSFW"功能的朋友只能上"Tensor Hub"去自由發揮了。
Additionally, the basic realistic model comes with the "NSFW" function, and due to the heavy merging ratio, the "NSFW" function is still retained. For those who prefer the "NSFW" function, you can only go to "Tensor Hub" to freely explore it.
這裡特別說明：使用Euler a採樣器雖沒有dpmpp_3m_sde與dpmpp_2m這2個採樣器畫質那麼高清,但它好處是畫面(像臉型與頭髮間)過渡很柔順、平滑，不會太銳利。
It should be noted here that although the Euler a sampler does not produce the same high-definition image as the dpmpp_3m_sde and dpmpp_2m samplers, its advantage is that the transitions in the image (such as between face shape and hair) are very smooth and not too sharp.
另外強烈推薦在低步數(10steps)以下不要使用dpmpp_3m_sde,因採樣器特性(步數不夠)還來不及收縮便出圖,容易造成手指等出現糨糊狀黏在一起,但若是只單純畫上半身肖像(不露手指) ,那便是首選,髮絲、臉部細節與皮膚,真的比Euler a採樣器好太多；當然高步數(15steps以上，最好是25steps以上)情況下就不用考慮這問題。
Additionally, it is strongly recommended not to use dpmpp_3m_sde for low step counts (below 10 steps). Due to the characteristics of the sampler (insufficient steps), it does not have time to shrink before outputting the image, which can easily cause fingers and other parts to stick together in a mushy manner. However, if you are simply drawing an upper-body portrait (without showing fingers), then it is the first choice. Hair strands, facial details, and skin are much better than the Euler a sampler. Of course, this problem does not need to be considered when drawing high step counts (15 steps or more, preferably 25 steps or more).
而dpmpp_2m採樣器則用在半身照上,它的畫質介於dpmpp_3m_sde與Euler a,在低步數畫中景時不會有dpmpp_3m_sde採樣器造成手指出現糨糊狀黏在一起,也不會像Euler a畫質稍差,而且姿勢多變,每種採樣器都有其亮點功能,多嘗試便知道如何在適當場景選擇採樣器。
The dpmpp_2m sampler is used for half-body photos. Its image quality is between dpmpp_3m_sde and Euler a. When drawing the mid-range shot at a low step count, the dpmpp_3m_sde sampler will not cause fingers to stick together like paste, and it will not have slightly inferior image quality like Euler a, and the postures are changeable. Each sampler has its own highlight function. Try it more and you will know how to choose the sampler in the appropriate scene.