使用介绍
Pony No Score v7 是一个突破性的版本,它释放了 Pony 的巨大潜力,标志着迄今为止最大的更新。V7 将 55 种风格集成到一个模型中,每种风格都有单独的触发关键字,类似于 Pony 之前为模型添加分数的方式。目标是创建一个能够产生各种艺术风格的模型,而不是仅限于难以控制的少数几种。用户可以应用单个风格或组合不同的风格来创造自己完全独特的艺术视野。一旦你隐藏了你的提示和数据,没有人会知道你结合了哪些风格关键字来制作如此独特的外观——它将是你的独有的。
使用模型非常简单,像乐谱一样,但又不像乐谱。其背后的想法是基于不同艺术家的各种艺术风格和一些独特的AI模型进行训练,将它们排列成单独的关键词(也许我应该将它们设为数字代码,就像Pony base中的score_1234,例如style_123——也许下次我会这样做)。每个关键词对应一种不同的风格,代表一位艺术家或一种独一无二的特殊AI风格。用户可以只调用一个关键词来调出那种风格,但这不是主要意图。我的目标是独特的风格组合。你可以一次调用多种风格,例如:A,B,C,或B,C,A,或C,A,B。
关键词或关键词顺序的任何变化都会产生不同的结果。你甚至可以使用括号(如(A)或(A:1.2))来增加或减少关键词的权重。这种组合和排列将使最初的 55 种风格变成数百或数千种不同的风格,只要你不透露你的风格组合和关键词,它就是你独一无二的作品。这是我创建这个项目的初衷。
说实话,完成这个模型后,我仍然发现一些错误,对结果并不完全满意。我相信如果重新开始,我可以取得更好的结果。但重做所有事情会花费太多时间,而且我没有足够的钱继续浪费资源。也许我必须等到下一个版本,届时我将从 55 种样式扩展到 100 种样式。
好的,以下是 55 种风格的完整列表:
触发词 | Arcane ttguweiz carmesjocelin fiz-rot anninosj samdoesarts elliemaplefox sei5osuzuki BM94199 yaoyao794 sea_Pall Vergilhoo hamukukka omao_51061954 letsfinalanswer 106858 sungmoomoo hakwansoo36 wonbin_lee hex_jjung Ghibli niji3d Hoovesart cutesexyrobutts afrobull unfairr Khyle shexyo Sinozick ironlilyart mossacannibalis gabrielpicolo Rossdraws Kuvshinov_Ilya RDJlock wlopwangling beitemian strauzek RyanReos hdfurrybabes Apone3D wildeerstudio AkiyamaRyo fugtrup crazytown 748cm ArawAraw stevencarson rescraft aromasensei ponyvid ponycos ghost236 7200ss lizchiefffff |
作品参照
source_anime, score_9, score_8_up, score_7_up, (masterpiece:1.2), (best quality:1.3), 1girl, thumbs up, tetradic Crimson theme, The Gilded Hollows background <lora:age_slider_v4:0.5>, low-key lighting, lo-fi, dark, glow, dynamic cinematic lighting, ray_tracing, global illumination <lora:Anime Style (Vauxz) PONY:0.8>
Negative prompt: low quality, score_6_down, score_5_down, score_4_down, score_3_down, score_2_down, score_1, source_pony, source_furry, femboy, low quality, low definition, NEGATIVE_HANDS, blurry, monochrome, frame, phone, Polaroid, picture frame, border, grainy, noisy, mutated hands and fingers, out of frame, extra limb, watermark, signature, bad anatomy, bad eyes, bad proportions, cropped, crossed eyes, dead eyes, deformed, disconnected limbs, disfigured, doll, extra fingers, extra limbs, floating limbs, long body, long neck, low-res, malformed hands, missing limb, mutated, mutated hands, mutation, noisy, out of focus, out of frame double, out of shot, poorly drawn face, poorly drawn hands, two faces, two heads, ugly, ugly eyes, watermark, text, cropped, bad art, deformed, worst quality, bad lighting, bad angle, boring, low-resolution, worst quality, bad composition, terrible lighting, holding paintbrush, easel, bad photo, boring, bad 3D render, children's crayon drawing, worst quality, blurry, blur, jpeg artifacts, compression artifacts, text, worst quality, noise, messy, low resolution, long torso, bright, spotlight, front lit, camera flash
Steps: 30, CFG scale: 7.5, Sampler: Euler a, Seed: 2674611862, TI: [object Object], Size: 832x1216, Model: DucHaiten Pony XL (no-score) v7.0, Version: f2.0.1v1.10.1-previous-564-g77464215, Module 1: ponyDiffusionV6XL_vae, Model hash: 1be0e3deca, Schedule type: Karras, ADetailer model: person_yolov8n-seg.pt, ADetailer version: 24.9.0, ADetailer mask blur: 4, ADetailer model 2nd: face_yolov8n_v2.pt, ADetailer model 3rd: full_eyes_detect_v1.pt, ADetailer model 4th: hand_yolov8s.pt, ADetailer confidence: 0.3, ADetailer prompt 4th: hand, ADetailer dilate erode: 4, ADetailer mask blur 2nd: 4, ADetailer mask blur 3rd: 4, ADetailer mask blur 4th: 4, ADetailer confidence 2nd: 0.3, ADetailer confidence 3rd: 0.3, ADetailer confidence 4th: 0.3, ADetailer inpaint padding: 32, ADetailer dilate erode 2nd: 4, ADetailer dilate erode 3rd: 4, ADetailer dilate erode 4th: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding 2nd: 32, ADetailer inpaint padding 3rd: 32, ADetailer inpaint padding 4th: 32, ADetailer denoising strength 2nd: 0.4, ADetailer denoising strength 3rd: 0.4, ADetailer denoising strength 4th: 0.4, ADetailer inpaint only masked 2nd: True, ADetailer inpaint only masked 3rd: True, ADetailer inpaint only masked 4th: True, Clip skip: 2
samdoesarts, anninosj, letsfinalanswer, hamukukka, elliemaplefox, sei5osuzuki, BM94199, yaoyao794, score_9, score_8_up, score_7_up, i want the whole image to be created in 3D anime style, solo, looking at viewer, short hair, bangs, simple background, black hair, 1boy, white background, hair between eyes, jewelry, male focus, earrings, mole, collar, grey eyes, black shirt, mole under eye, chain, scar, mouth hold, portrait, scar on face, close-up, metal collar
Negative prompt: score_6, score_5, score_4, lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, bad art, bad painting, bad photo, bad image
Steps: 30, CFG scale: 9, Sampler: Euler a, Seed: 1848536828, RNG: CPU, Size: 720x1280, Model: Pony-no-score-v7-7-step00019740, Version: v1.10.1, Model hash: 1be0e3deca, Schedule type: SGM Uniform, Clip skip: 2
score_9, score_8_up, score_7_up, i want the whole image to be created in 3D anime style, 1girl, solo, looking at viewer, short hair, bangs, shirt, black hair, brown eyes, jewelry, closed mouth, collarbone, white shirt, upper body, short sleeves, outdoors, glasses, day, necklace, blurry, lips, depth of field, blurry background, messy hair, round eyewear
Negative prompt: score_6, score_5, score_4, lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, bad art, bad painting, bad photo, bad image
Steps: 30, CFG scale: 9, Sampler: Euler a, Seed: 3419531008, RNG: CPU, Size: 720x1280, Model: Pony-no-score-v7-7-step00019740, Version: v1.10.1, Model hash: 1be0e3deca, Schedule type: SGM Uniform, Clip skip: 2
使用介绍
添加了 pony-no_score 版本,它基本上是 pony 模型,但在没有 tag_score 的情况下可以保持 70-90% 的图像质量,当然,如果它仍然有 tag_score 会更好
特别提醒:pony base v6 版似乎对某些关键字或字符(单独或组合)有错误,在生成图片时会导致各种严重错误。这是 pony base 的错误。希望 pony base v7 版能改善这个问题,否则以后开发和扩展会非常困难,虽然我已尽力改善。
因为civitai和automatic1111的图像生成器不同,所以在civitai上需要将步数设置得比使用automatic1111时高大约10-15步。
2024年9月8日更新:
终于赶上了 pony base v6 LOL。除了修复错误之外,这可能是最后一次更新了。我觉得 pony v6 已经达到了极限,再添加任何东西都可能导致其他地方出现问题。
6.0 版本包含一些动漫数据、针对 NSFW 的更好调整、对颜色和灯光效果的进一步调整以及对自然语言的额外改进。
基于原始pony模型并结合我的各种模型创建的模型。我尽量保留了 pony 的几乎所有文本编码器,只更改了图像数据,所以如果你不知道如何正确使用“标签分数”,我建议你多了解一下 pony。使用这个模型,你会觉得图像的随机性更小,没有像 pony 那样有太多不必要的细节,而且会感觉更平滑、更干净,消除了某些提示会产生刺眼图像或烧焦颜色的可能性,脸部变得更清晰,手、脚和姿势调整后错误更少。但得其一,失其二。只能说,每个人都有自己的风格,看你的喜好。
作品参照
1880s adult Victorian wife mother with long blonde hair worn out and full, wearing light clothes, full lips and a soft face and large green eyes, score_9, score_8_up, score_7_up
Negative prompt: score_6, score_5, score_4, lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, bad art, bad painting, bad photo, bad image
Steps: 30, CFG scale: 9, Sampler: Euler a, Seed: 2437079973, RNG: CPU, Size: 720x1280, Model: DucHaiten-Pony-No_Score_v6.0, Version: v1.10.1, Mask blur: 4, Model hash: 354679d31b, Inpaint area: Only masked, Hires upscale: 2, Schedule type: SGM Uniform, Hires upscaler: 4xUltrasharp_4xUltrasharpV10, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
score_9, score_8_up, score_7_up, 1girl, solo, looking at viewer, short hair, blue eyes, blonde hair, simple background, jewelry, closed mouth, earrings, sleeveless, hair over one eye, sweater, lips, makeup, turtleneck, portrait, realistic, nose, android 18
Negative prompt: score_6, score_5, score_4, lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, bad art, bad painting, bad photo, bad image
Steps: 30, CFG scale: 9, Sampler: Euler a, Seed: 470916059, RNG: CPU, Size: 720x1280, Model: DucHaiten-Pony-No_Score_v6.0, Version: v1.10.1, Mask blur: 4, Model hash: 354679d31b, Hires prompt: [object Object], Inpaint area: Only masked, Hires upscale: 2, Schedule type: SGM Uniform, Hires upscaler: 4xUltrasharp_4xUltrasharpV10, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
score_9, score_8_up, score_7_up, 1girl, solo, breasts, short hair, thighhighs, gloves, dress, cleavage, bare shoulders, medium breasts, sitting, white hair, hairband, boots, outdoors, black gloves, elbow gloves, black thighhighs, water, mole, black dress, clothing cutout, thigh boots, black hairband, plant, building, mole under mouth, rain, blindfold, ruins, covered eyes, black blindfold, overgrown, post-apocalypse, yorha no. 2 type b
Negative prompt: score_6, score_5, score_4, lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, bad art, bad painting, bad photo, bad image
Steps: 30, CFG scale: 9, Sampler: Euler a, Seed: 3132262904, RNG: CPU, Size: 720x1280, Model: DucHaiten-Pony-No_Score_v6.0, Version: v1.10.1, Mask blur: 4, Model hash: 354679d31b, Hires prompt: [object Object], Inpaint area: Only masked, Hires upscale: 2, Schedule type: SGM Uniform, Hires upscaler: 4xUltrasharp_4xUltrasharpV10, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
使用介绍
添加了 pony-no_score 版本,它基本上是 pony 模型,但在没有 tag_score 的情况下可以保持 70-90% 的图像质量,当然,如果它仍然有 tag_score 会更好
特别提醒:pony base v6 版似乎对某些关键字或字符(单独或组合)有错误,在生成图片时会导致各种严重错误。这是 pony base 的错误。希望 pony base v7 版能改善这个问题,否则以后开发和扩展会非常困难,虽然我已尽力改善。
因为civitai和automatic1111的图像生成器不同,所以在civitai上需要将步数设置得比使用automatic1111时高10-15步左右。
2024年8月8日更新:
v5.1+v5.0+v4.0=v5.2
该死,我忘了开启抢先体验功能了
版本 v5.2 无论有无分数都运行良好
作品参照
Close-up details, realistic style photo Japanese young woman, black hair, long hair, outstanding style, tall, cute, in her teens. Use soft lighting to cast gentle shadows on the subject, adding a touch of dimension to your images without compromising details, score_9, score_8_up, score_7_up
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, CFG scale: 9, Sampler: Euler a, Seed: 3907219628, RNG: CPU, Size: 720x1280, Model: DucHaiten-Pony-No_Score_v5.2, Version: v1.10.1, Mask blur: 4, Model hash: ee591aba60, Inpaint area: Only masked, Hires upscale: 2, Schedule type: SGM Uniform, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
score_9, score_8_up, score_7_up, 1girl, solo, breasts, short hair, dress, bare shoulders, medium breasts, closed mouth, upper body, white hair, hairband, sleeveless, mole, blurry, black dress, lips, wet, depth of field, blurry background, turtleneck, phone, cellphone, black hairband, wet clothes, mole under mouth, facing viewer, smartphone, rain, water drop, blindfold, wet hair, covered eyes, black blindfold, yorha no. 2 type b
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, CFG scale: 9, Sampler: Euler a, Seed: 2882857081, RNG: CPU, Size: 720x1280, Model: DucHaiten-Pony-No_Score_v5.2, Version: v1.10.1, Mask blur: 4, Model hash: ee591aba60, Inpaint area: Only masked, Hires upscale: 2, Schedule type: SGM Uniform, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
score_9, score_8_up, score_7_up, 1girl, solo, short hair, blue eyes, black hair, long sleeves, japanese clothes, barefoot, kimono, sash, obi, bubble, white kimono, underwater, air bubble, submerged, ryougi shiki
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, CFG scale: 9, Sampler: Euler a, Seed: 2055059879, RNG: CPU, Size: 720x1280, Model: DucHaiten-Pony-No_Score_v5.2, Version: v1.10.1, Mask blur: 4, Model hash: ee591aba60, Hires prompt: [object Object], Inpaint area: Only masked, Hires upscale: 2, Schedule type: SGM Uniform, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
使用介绍
2024年7月23日更新:
欢迎来到 Pony No Score v5。我试图将它保存到 Pony Base v6.9 发布,但等待的时间太长了,所以我不能再等了,立即开始工作。我需要它有更多的灵魂,更多的艺术深度,并在每幅创作的图像中传达情感。我受够了人工智能生成的艺术的枯燥,它只看起来很美,但感觉很空洞,不像真正从创作者的灵魂中培育出来的东西。我不知道你是否明白我的意思,但这就是我想要实现的。也许对你来说最明显的变化是 3D 风格将被最小化,2D 绘图风格将经历轻微的修改,颜色感觉更具电影感,灯光和阴影将更好更逼真。
其他一些更新:改进的文本编码器、更多的拍摄角度、更多的艺术和 2D 数据来取代 Nijijourney 数据
推荐设置:
采样方法:DPM++ SDE
时间表类型:SGM Uniform(此项为必填项)
采样步骤:30-36
CFG Scale:9
添加了 pony-no_score 版本,它基本上是 pony 模型,但在没有 tag_score 的情况下可以保持 70-90% 的图像质量,当然,如果它仍然有 tag_score 会更好
特别提醒:pony v6 base 似乎对某些关键字或字符(单独或组合)有错误,在生成图像时会导致各种严重错误。这是 pony base 的错误。希望 pony base v7 可以改善这个问题,否则以后开发和扩展会非常困难,虽然我做了很多改进。
因为civitai和automatic1111的图像生成器不同,所以在civitai上需要将步数设置得比使用automatic1111时高大约10-15步。
作品参照
score_9, score_8_up, score_7_up, 1girl, solo, breasts, short hair, dress, bare shoulders, medium breasts, closed mouth, upper body, white hair, hairband, sleeveless, mole, blurry, black dress, lips, wet, depth of field, blurry background, turtleneck, phone, cellphone, black hairband, wet clothes, mole under mouth, facing viewer, smartphone, rain, water drop, blindfold, wet hair, covered eyes, black blindfold, yorha no. 2 type b
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, RNG: CPU, Size: 720x1280, Seed: 3001677365, Model: DucHaiten-Pony-no_score_v5.0, Version: v1.9.4, Sampler: DPM++ SDE, CFG scale: 9, Clip skip: 2, Mask blur: 4, Model hash: b5bda9465c, Inpaint area: Only masked, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True
score_9, score_8_up, score_7_up, 1girl, solo, long hair, breasts, looking at viewer, blush, smile, bangs, gloves, hat, twintails, medium breasts, swimsuit, white hair, cowboy shot, outdoors, parted lips, horns, day, elbow gloves, pink eyes, water, bag, blurry, lips, wet, one-piece swimsuit, covered navel, blurry background, headphones, white headwear, highleg, ass visible through thighs, backpack, wet clothes, clothes writing, baseball cap, skin tight, wading, competition swimsuit, white one-piece swimsuit, highleg swimsuit, pink gloves, wet swimsuit, pink one-piece swimsuit
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, RNG: CPU, Size: 720x1280, Seed: 3001677370, Model: DucHaiten-Pony-no_score_v5.0, Version: v1.9.4, Sampler: DPM++ SDE, CFG scale: 9, Clip skip: 2, Mask blur: 4, Model hash: b5bda9465c, Inpaint area: Only masked, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True
score_9, score_8_up, score_7_up, i want the whole image to be created in 3D anime style, solo, looking at viewer, blue eyes, gloves, 1boy, holding, closed mouth, weapon, white hair, male focus, japanese clothes, black gloves, sword, fingerless gloves, holding weapon, blurry, muscular, blurry background, facial hair, holding sword, katana, sheath, realistic, manly
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, RNG: CPU, Size: 720x1280, Seed: 3001677360, Model: DucHaiten-Pony-no_score_v5.0, Version: v1.9.4, Sampler: DPM++ SDE, CFG scale: 9, Clip skip: 2, Mask blur: 4, Model hash: b5bda9465c, Inpaint area: Only masked, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True
使用介绍
添加了 pony-no_score 版本,它基本上是 pony 模型,但在没有 tag_score 的情况下可以保持 70-90% 的图像质量,当然,如果它仍然有 tag_score 会更好
特别提醒:pony v6 base 似乎对某些关键字或字符(单独或组合)有错误,在生成图像时会导致各种严重错误。这是 pony base 的错误。希望 pony base v7 可以改善这个问题,否则以后开发和扩展会非常困难,虽然我做了很多改进。
因为civitai和automatic1111的图像生成器不同,所以在civitai上需要将步数设置得比使用automatic1111时高大约10-15步。
2024 年 5 月 27 日更新:这将是 pony base v7 发布前的最终版本。
基于原始pony模型并结合我的各种模型创建的模型。我尽量保留了 pony 的几乎所有文本编码器,只更改了图像数据,所以如果你不知道如何正确使用“标签分数”,我建议你多了解一下 pony。
使用这个模型,你会觉得图像的随机性更小,没有像 pony 那样有太多不必要的细节,而且会感觉更平滑、更干净,消除了某些提示会产生刺眼图像或烧焦颜色的可能性,脸部变得更清晰,手、脚和姿势调整后错误更少。但得其一,失其二。只能说,每个人都有自己的风格,看你的喜好。
作品参照
score_9, score_8_up, score_7_up, i want the whole image to be created in 3D anime style, 1girl, solo, looking at viewer, short hair, bangs, shirt, black hair, brown eyes, jewelry, closed mouth, collarbone, white shirt, upper body, short sleeves, outdoors, glasses, day, necklace, blurry, lips, depth of field, blurry background, messy hair, round eyewear
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, RNG: CPU, Size: 768x1128, Seed: 1684575564, Model: DucHaiten-Pony-no_score_v4.0, Version: v1.9.3, Sampler: DPM++ 2M, CFG scale: 8, Clip skip: 2, Mask blur: 4, Model hash: 886a42c049, Hires prompt: [object Object], Inpaint area: Only masked, Hires upscale: 2, Schedule type: Karras, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True
score_9, score_8_up, score_7_up, Girl's profile picture, light green long hair with bangs, light blue petals on cheeks, realistic skin texture, detailed picture, close-up, HD32k
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, : [object Object], RNG: CPU, Size: 768x1128, Seed: 3478070916, Model: DucHaiten-Pony-no_score_v4.0, Version: v1.9.3, Sampler: DPM++ 2M, CFG scale: 8, Clip skip: 2, Mask blur: 4, Model hash: 886a42c049, Hires prompt: [object Object], Inpaint area: Only masked, Hires upscale: 2, Schedule type: Karras, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True
score_9, score_8_up, score_7_up, i want the whole image to be created in 3D anime style, solo, simple background, shirt, yellow eyes, upper body, vest, ascot, no humans, floral print, formal, brown background
Negative prompt: score_6, score_5, score_4, source_pony, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs, painting by bad-artist
Steps: 30, RNG: CPU, Size: 768x1128, Seed: 1684575561, Model: DucHaiten-Pony-no_score_v4.0, Version: v1.9.3, Sampler: DPM++ 2M, CFG scale: 8, Clip skip: 2, Mask blur: 4, Model hash: 886a42c049, Hires prompt: [object Object], Inpaint area: Only masked, Hires upscale: 2, Schedule type: Karras, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8n.pt, ADetailer version: 23.7.11, Denoising strength: 0.4, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True