GonzLomo 女性自拍写真模型,是一款精心打造且独具特色的基于sdxl训练的大模型。它以先进的Stable Diffusion XL为基石,通过对大量女性自拍相关素材的深度学习与训练,优化了U – Net、VAE、CLIP Text Encoder等关键组件,能精准捕捉女性自拍时的神态、姿态与光影效果,生成极为逼真且具艺术感的女性自拍摄影风格图像 。
关于v2.0 Unity DMD版本
本质上,这是 V1.1 Real,并添加了 Siren Forge、Cyberrealistic 和 Lustify 的元素。基础仍然是 Gonzales 和 LomoXL。
我把这个版本命名为 Unity,是因为我最初想为照片级写实风格和艺术风格分别设计模型。然而,我意识到可以把两者整合到一个模型中,并为此创建了 Artsy 1.1 版本,但最终失败了。这是我第二次尝试这种组合,我认为/希望它会成功。
至少,我不想降低照片写实能力,只是想添加更多超现实、艺术或幻想风格作为选项。
自 V1.1 Real 以来,推荐的设置没有改变:
采样器:LCM(dpmpp2a 和 ddim 也可以提供有趣的结果)
调度器:Karras、Exponential 或 Beta
steps:8-12
CFG:1.0-1.3
Clip Skip:1-5
我通常将设置保持在 LCM/Karras 格式,CFG 设为 1.0,但我倾向于将Clip Skip值在 1 到 5 之间变化,因为这些值可能会对图像产生相当大的影响。有时,Clip Skip值 2 的图像效果会很糟糕,而Clip Skip 1 的图像可能会变得非常棒,反之亦然。
我相信这个模型在照片写实方面非常出色,尽管它也具备一些艺术性。以下是一些充分利用GonzaLomo潜力的技巧:
- Clip Skip是你的好帮手。我建议尝试同时使用 1 和 2 的提示。1 可以拍出更具创意、艺术感、复古或粗犷风格的图像,而 2 则更注重照片写实感,以及主体的魅力。
- 使用ComfyUI中的PatchModelAddDownscale节点生成分辨率为 1024×1536 或更高的图像。这将创建细节更丰富的图像,之后可以进一步放大,从而获得令人难以置信的照片级逼真细节。使用该节点,我可以进行高达 1280×1920 的初始生成。
- 您可以尝试不同的采样器设置,但我发现绝对最好的结果是使用 LCM/Karras、12 步和 CFG 1.0。
- 选择性地使用Self Attention Guidance节点有时能带来意想不到的效果。我总是把它放在流程中,通常情况下会绕过它,但当我觉得图像有潜力,但只需要一些细节提升,比如色彩饱和度或对比度时,我就会启用它。
- 我使用V1.1 半真实模型生成图像,然后用V2.0 Unity模型以大约 0.60 的降噪值重新采样,得到了非常有趣/富有创意的结果。你确实需要反复调整降噪值,才能找到合适的值,让图像从卡通风格过渡到照片级写实风格,但当你这样做时,你会惊讶于它所能带来的效果。需要注意的是,V1.1 半真实模型需要 Clip Skip 2+ 以上版本,1 版本完全无法使用。
- 最后,一定要不断升级!像素越高,细节越多,照片越逼真。
注意:我只使用 ComfyUI,因此如果您在其他工具中使用我的模型时遇到问题而我无法提供帮助,我提前表示歉意。
作品参照

cinematic still score_9, score_8_up, score_7_up, (realistic:1.2), realistic, photo, aesthetic erotic photography . emotional, harmonious, vignette, 4k epic detailed, shot on kodak, 35mm photo, sharp focus, high budget, cinemascope, moody, epic, gorgeous, film grain, grainy, (masterpiece), (best quality), (ultra-detailed), score_9, score_8_up, score_7_up, (realistic:1.2), realistic, photo, aesthetic erotic photography, illustration, disheveled hair, detailed eyes, perfect composition, moist skin, intricate details, earrings, source_hustler,raw,photo,amateur,1girl,age24,(long hair, bright blonde hair, blue eyes, smile, Au naturel, realistic skin, photo realistic, sharp iris, eye lashes, black mini winged eyeliner:a thin line of eyeliner just along the top lash line extending slightly at the outer corner to create a small flick, realistic pupils, punk rock attitude),ponytail,crop top,dressed in daisy dukes,sassy,medium breasts,underboob nipple slip,see through nipples, Hyperrealistic art cinematic film still photography in the style of detailed hyperrealism photoshoot,RAW,color graded portra 400 film,remarkable color,realistic dull skin noise,Thorough,analog style,highest quality,skin pores,sharp focus,dappled lighting,film grain,photographed on a Sony A7R IV,18mm F/1.7 cine lens,(highly detailed, intricately detailed),8k,HDR, masterpiece, best quality, newest, late,ultrasharp, score_9, score_8_up, score_7_up, (realistic:1.2), realistic, photo, aesthetic erotic photography, intricate, very beautiful, highly detailed, extremely detail, sacred, enigmatic, lucid, glowing, magical, sharp focus, winning, illuminated, cinematic, artistic, fine classic, grand elaborate excellent, perfect composition, cute
Negative prompt: text, motion lines, effects, border, frame.
(worst quality, low quality, normal quality, lowres, low details, oversaturated, undersaturated, overexposed, underexposed, grayscale, bad photo, bad photography, bad art:1.4)
Steps: 10, CFG scale: 1, Sampler: LCM, Seed: 286, Model: gonzaLomo-v11-40_sirenForge-v20-30_cyber-v56-30_00001_

score_9, score_8_up, score_7_up, Anime Style, Cowboy Shot, girl, tunisian, cute, seductive, innocent, light smile:0.3, plump lips, slender body, short hair, slender, red lips, wearing glasses, wearing high detailed red dress, sitting in a lounge café, cozy atmosphere, flirting, soft light passing through hair, side lighting, reflected light, depth of field, dynamic angle, fashion photography, sharp, hyperdetailed:1.15, [red pubic hair]
Negative prompt: text, motion lines, effects, border, frame.
(worst quality, low quality, normal quality, lowres, low details, oversaturated, undersaturated, overexposed, underexposed, grayscale, bad photo, bad photography, bad art:1.4)
Steps: 9, CFG scale: 1, Sampler: LCM, Seed: 1147, Model: DMD\GB-Mix_DMD_v20_Unity

1girl, solo, 26 year old, selfie, amateur photo, dark brown eyes, looking at viewer, fair skin, pale skin, skinny, closed mouth smile, dark blonde hair, (wavy hair:0.7), double bun, eyeliner, (yellow sundress with white floral pattern:1.2), cleavage, legs, thighs, sitting in chair, bedroom
Negative prompt: infant, underage, young, child, teen, bad quality, worst quality, low quality, worst detail, deformed, bad hands, bad eyes, signature, username, logo, watermark, text, label, cartoon, illustration, anime, painting, CGI, 3D render
Steps: 10, CFG scale: 1.1, Sampler: LCM, Seed: 1529985136, Size: 832x1280, Model: gonzalomo_v20UnityDMD, Version: f2.0.1v1.10.1-previous-659-gc055f2d4, Model hash: 4a2a8fe0f7, Schedule type: Exponential, Clip skip: 2