SD生成的真人看的多了实际感觉有点千篇一律，越来越感觉不够真实，比P的图还假。AI脸看起来以后也是一个专门的分类了。
不过在C站上看到一个很有趣的LORA模型 “墨心” 这就很中国风了。尤其油画风格加上墨心就特别有意思了。
我也试试生成了几张确实质量很高。
可以用来做什么呢，目前我觉得很有前景的一个应用是给小说或者文章配图。
生成参数如下：
(ultra high res,photorealistic,realistic,best quality,photo-realistic),
(high detailed skin,visible pores),(real person,photograph),
(((pink hanfu,chinese style))),realistic,simple background,upper body,
(8k, raw photo, best quality, masterpiece),(1girl),photon mapping, radiosity, physically-based rendering,automatic white balance,(((haunting smile,long hair))),watery eyes,
(blush|cute and playful|adorable|skinny|thick bangs|beauty|twintail|hair bun,),(looking at viewer),(irises and pupils are rounded,the pupil reflects the surroundings,eyes are not the same size), <lora:Moxin_10:0.4>
Negative prompt: nsfw,(worst quality:1.7),(low quality:1.6),(normal quality:1.5),lowres,blurry,simple background,jpeg artifacts,bad-artist,bad shadow,compressed image,low pixel,light spot,
paintings,sketches,((monochrome)),((grayscale)),paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2), lowres, normal quality, ((monochrome)), ((grayscale)), skin spots, acnes, skin blemishes, age spot, glans
Steps: 28, Sampler: DPM++ SDE Karras, CFG scale: 8, Seed: 3102037116, Size: 512x624, Model hash: 7107c05c1c, Model: dalcefoPainting_3rd, Denoising strength: 0.7, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent
SD的缺点还是把握不住手部绘制。
