Pony Realism 真实大模型

Pony Realism真实大模型是一款别具一格且功能强大的图像生成模型，它基于pony精心训练而成。在C站广受欢迎的pony模型基础上，Pony Realism真实模型通过对大量数据的深度学习，拥有强大的图像生成能力，能依据各类提示词，生成细节丰富、风格多样的图像，无论是逼真的人物形象，还是奇幻的场景画面，它都能出色呈现。

第二页包括Pony Realism 模型详细使用指南

关于v2.3版本

主版本：本次更新主要是为了保持画质的更新。它改善了光照效果（更自然，暗部减少），使面部和皮肤看起来更干净，整体图像更加锐利，并更好地展现细微的细节。

• 改善了照明，使其更加自然，减少了黑暗。
• 增强整体图像质量和清晰度。
• 更加精致和细致的面部。
• 皮肤更光滑、更逼真。
• 细微细节更加清晰。

🏠本地生成的推荐参数

采样器：

• 受到推崇的：
  • DPM++ 2M SDE Karras
  • DPM++ 2S a Karras
  • DPM++ SDE Karras
  • DPM++ 2M SDE Exponential
  • DPM2 a
  • DPM++ 2S a
  • DPM++ 3M *
  • Euler A

设置：

• Steps： >=30
• CFG：6-7
• Clip Skip： 2
• 分辨率： > 1024
• 术语：建议使用“female/male”而不是“woman/man”

附加信息：

• 使用Danbooru标签。
• 提示的权重不要超过1.5
• 提示样式：
  • 积极提示： score_9, score_8_up, score_7_up, BREAK
  • 消极提示： score_4, score_5, score_6

作品参照

score_9, score_8_up, score_7_up, forest, plant close-up, water drops, highly detailed, film grain, DSLR photo, best quality, 85mm lens, shallow depth of field, soft lighting, highly detailed, cinematic composition, Product-View
Negative prompt: score_6, score_5, score_4, text, censored, deformed, bad hand, watermark
Steps: 35, CFG scale: 6, Sampler: DPM++ 3M SDE, Seed: 1452109464, Size: 832x1216, Model: PonyRealism_v2.3_VAE, Version: f1.2.5-v1.10.1RC-latest-983-g44e5eb17, Model hash: b4d6dee26f, Schedule type: Polyexponential, Clip skip: 2

score_9, score_8_up, score_7_up, young man, 1800s, brown suit, watch, firm face, highly detailed, high contrast, film grain, superabsurd res, 8k
Negative prompt: score_6, score_5, score_4, text, censored, deformed, bad hand, watermark
Steps: 35, CFG scale: 3, Sampler: DPM++ 3M SDE, Seed: 2605886797, Size: 832x1216, Model: PonyRealism_v2.3_VAE, Version: f1.2.5-v1.10.1RC-latest-983-g44e5eb17, Model hash: b4d6dee26f, Schedule type: Polyexponential, ADetailer model: person_yolov8s-seg.pt, ADetailer version: 24.11.1, ADetailer mask blur: 4, ADetailer model 2nd: yolov10m-face.pt, ADetailer model 3rd: hand_yolov8n.pt, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer mask blur 2nd: 4, ADetailer mask blur 3rd: 4, ADetailer confidence 2nd: 0.3, ADetailer confidence 3rd: 0.3, ADetailer inpaint padding: 32, ADetailer dilate erode 2nd: 4, ADetailer dilate erode 3rd: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding 2nd: 32, ADetailer inpaint padding 3rd: 32, ADetailer denoising strength 2nd: 0.4, ADetailer denoising strength 3rd: 0.19, ADetailer inpaint only masked 2nd: True, ADetailer inpaint only masked 3rd: True, Clip skip: 2

score_9, score_8_up, score_7_up, blonde hair, wavy hair, bedroom eyes, red lipstick, portrait, skimpy, Sequin Bra Top, (clutter:0.3), supported on bedside table, solo, curvy, busty, large breasts, looking at viewer, soft lighting, glamour , portrait, highly detailed, high contrast, film grain, superabsurd res, 8k
Negative prompt: score_6, score_5, score_4, text, censored, deformed, bad hand, watermark
Steps: 35, CFG scale: 5, Sampler: DPM2 a, Seed: 3797450829, Size: 832x1216, Model: PonyRealism_v2.3_VAE, Version: f1.2.5-v1.10.1RC-latest-983-g44e5eb17, Model hash: b4d6dee26f, Schedule type: Karras, ADetailer model: person_yolov8s-seg.pt, Wildcard prompt: [object Object], ADetailer version: 24.11.1, ADetailer mask blur: 4, ADetailer model 2nd: yolov10m-face.pt, ADetailer model 3rd: hand_yolov8n.pt, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer mask blur 2nd: 4, ADetailer mask blur 3rd: 4, ADetailer confidence 2nd: 0.3, ADetailer confidence 3rd: 0.3, ADetailer inpaint padding: 32, ADetailer dilate erode 2nd: 4, ADetailer dilate erode 3rd: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding 2nd: 32, ADetailer inpaint padding 3rd: 32, ADetailer denoising strength 2nd: 0.4, ADetailer denoising strength 3rd: 0.19, ADetailer inpaint only masked 2nd: True, ADetailer inpaint only masked 3rd: True, Clip skip: 2

使用介绍

✨ v2.2 就在这里！

此次更新在某些方面进行了小幅改进，在保留模型当前优势的同时进行了细微的改进。

具体变化如下：

• 改善照明
• 黑暗/夜间场景下有更好的表现
• 改善手脚
• 色彩更加精致
• 更清晰的整体图像质量
• 从不同角度看脸型略好一些
• 皮肤细微改善
• 使用 Euler A 增强输出
• 提高即时理解能力
• 细节更精致
• 更新的 VAE

提示和生成技巧

我写了一篇文章，解释并提供了一些有用的资源，以帮助您实现出色的生成结果。您可以查看它以获取详细的见解和建议。

🏠本地生成的推荐参数

采样器：

• 受到推崇的：
  • DPM++ 2M SDE Karras
  • DPM++ 2S a Karras
  • DPM++ SDE Karras
  • DPM++ 2M SDE Exponential
  • DPM2 a
  • DPM++ 2S a
  • DPM++ 3M *
  • Euler A

设置：

• Steps： >=30
• CFG：6-7
• Clip Skip： 2
• 分辨率： > 1024
• 术语：建议使用“female/male”而不是“woman/man”

附加信息：

• 使用Danbooru标签。
• 提示的权重不要超过1.5
• 提示样式：
  • 积极提示： score_9, score_8_up, score_7_up, BREAK
  • 消极提示： score_4, score_5, score_6

作品参照

score_9, score_8_up, score_7_up, 20 years old girl, shy, long hair, alternative vibe, beautiful eyes, freckles, medium breasts, subtle cleavage, finger shh, slutty, relaxing, candid picture, playful, foot focus, on front, bedroom, depth of field, highly detailed, high contrast, film grain, Rim Lighting, Rays of Shimmering Light, low light
Negative prompt: score_6, score_5, score_4, text, censored, deformed, bad hand
Steps: 30, CFG scale: 6, Sampler: DPM++ SDE, Seed: 3690904707, VAE: sdxl.vae.safetensors, Size: 832x1216, Model: PonyRealism_v2.2_VAE, Version: v1.10.1, Template: [object Object], Model hash: 7c97ecf786, Schedule type: Karras, Variation seed: 2501343776, ADetailer model: mediapipe_face_full, ADetailer version: 24.8.0, Negative Template: [object Object], ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, Downcast alphas_cumprod: True, Variation seed strength: 0.07, 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, female, succubus, portrait, sit in a log, long tongue, holding wine glass, red theme, skimpy dress, horns, tail, fire, wasteland, medieval fantasy, ash, depth of field, highly detailed, high contrast, film grain, Rim Lighting
Negative prompt: score_6, score_5, score_4, text, censored, deformed, bad hand
Steps: 30, CFG scale: 6, Sampler: DPM++ SDE, Seed: 2985911117, VAE: sdxl.vae.safetensors, Size: 832x1216, Model: PonyRealism_v2.2_VAE, Version: v1.10.1, Template: [object Object], Model hash: 7c97ecf786, Schedule type: Karras, ADetailer model: mediapipe_face_full, ADetailer version: 24.8.0, Negative Template: [object Object], ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, Downcast alphas_cumprod: True, 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, scene, huge castle, moss, mist, fog, mistery, dark fantasy, low light, depth of field, highly detailed, high contrast, film grain, Rim Lighting, Long Exposure, DSLR
Negative prompt: score_6, score_5, score_4, text, censored, deformed, bad hand
Steps: 30, CFG scale: 6, Sampler: DPM++ SDE, Seed: 4268524215, VAE: sdxl.vae.safetensors, Size: 832x1216, Model: PonyRealism_v2.2_VAE, Version: v1.10.1, Template: [object Object], Model hash: 7c97ecf786, Schedule type: Karras, ADetailer model: mediapipe_face_full, ADetailer version: 24.8.0, Negative Template: [object Object], ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, Downcast alphas_cumprod: True, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2

使用介绍

v2.1是对模型的一次重大改造。

此版本更倾向于 Pony，以获得更多整体变化，但保持了逼真的质量。示例图像没有Adetailer

推荐参数：

·采样器：

1. DPM++ SDE Karras
2. DPM++ 2M Karras
3. DPM++ 2M SDE Karras
4. DPM++ 2M SDE Exponential

我不建议使用以下采样器（细节较弱）：

• Euler a
• Euler

步骤： 20 – 30。

CFG：6-7

剪辑跳过：2

分辨率：我一直使用1280×768 | 768×1280

·我建议使用female/male而不是woman/man

此版本不需要任何触发词，并且与大多数 Pony 模型一样，它使用以下提示样式：

积极提示：

score_9, score_8_up, score_7_up, BREAK

消极提示：

score_1, score_2, score_3

💪该模型具有更普遍的多样性、创造性和更好的即时依从性。

如果距离较远，建议使用ADETAILER 。

💕这可能是目前模型的最后一个版本，因为我对结果非常满意。感谢您下载，希望您喜欢它 :)

作品参照

Stable Diffusion提示词

score_9, score_8_up, score_7_up BREAK 1girl, solo, long hair, ginger hair, grey eyes, fox ears, black t-shirt, large breasts,
Negative prompt: score_4, score_5, score_6,
Steps: 30, Seed: 959079650863927, Model: NSFW\SDXL\ponyRealism_v21VAE, width: 1024, height: 1024, Sampler: dpmpp_3m_sde_gpu, CFG scale: 4.5, modelIds: , versionIds: 

cinematic film still cinematic photo breathtaking score_9, score_8_up, score_7_up, BREAK, 8k, masterpiece, high quality, long hair, necklace, hoop earrings, thigh strap, shirt and jeans, skin pores, laying in a confy chair, garden, highly detailed, detailed skin, depth of field, film grain, highly detailed, detailed skin, depth of field, film grain, <lora:CinematicStyle_v1:1> . award-winning, professional, highly detailed . 35mm photograph, film, bokeh, professional, 4k, highly detailed . shallow depth of field, vignette, highly detailed, high budget Hollywood movie, bokeh, cinemascope, moody, epic, gorgeous, film grain, grainy
Negative prompt: score_1, score_2, score_3, text, ugly, deformed, noisy, blurry, distorted, grainy, drawing, painting, crayon, sketch, graphite, impressionist, noisy, blurry, soft, deformed, ugly, anime, cartoon, graphic, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured
Steps: 30, VAE: sdxl_vae.safetensors, Size: 768x1280, Seed: 621014382, Model: PonyRealism_v2.1_VAE, Version: 1.7.0, Sampler: DPM++ SDE Karras, Template: [object Object], CFG scale: 7, Clip skip: 2, Model hash: 059934ff58, ADetailer model: mediapipe_face_full, ADetailer version: 24.1.2, Negative Template: [object Object], ADetailer mask blur: 4, 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, BREAK, woman, bikini, skin pores, little smirk, sunglasses, pool, sunny, 80s vibe, depth of field, film grain
Negative prompt: score 1, score 2, score 3, text, penetration
Steps: 30, VAE: sdxl_vae.safetensors, Size: 768x1280, Seed: 3188932120, Model: PonyRealism_v2.1_VAE, Version: 1.7.0, Sampler: DPM++ SDE Karras, Template: [object Object], CFG scale: 6, Clip skip: 2, Model hash: 059934ff58, Negative Template: [object Object]

使用介绍

V1.1：整体真实感和质量的改进。

接下来的更新将集中在面孔多样性上，因为在我看来，真实感和质量都得到了极大的实现

从现在开始，模型将内置VAE

使用方法

CLIP SKIP：2
该版本需要如下提示：

• 积极的：

score_9, score_8_up, score_7_up, BREAK, source_real, raw, photo, amateur

• 消极的：

score_1, score_2, score_3, source_western, source_comic, source_furry

即使面孔得到了很大改善，我还是建议对面孔添加一些描述，以获得更多变化。

💡我建议使用下载地址里的wildcard

📍个人最喜欢的采样器是DPM++ 2M SDE Exponential

作品参照

Stable Diffusion提示词

score_9, score_8_up, score_7_up, BREAK, source_real, raw, photo, amateur, woman, Diamond Face, Close-Set Eyes, [eyecolors violet], turned up nose-shape, full lips, high cheekbones, weak receding chin, burgundy, lob, light blue, lip gloss, __15JeweleryMaterials__ __14Piercing__, Small and droopy breasts, tired, hot, gorgerous, big breasts, outdoor, cleavage, portrait, lust, highly detailed, detailed skin, depth of field, film grain
Negative prompt: score 1, score 2, score 3, source_western, source_comic, source_furry, worst quality, low quality, text, censored, deformed, bad hand
Steps: 40, Size: 768x1280, Seed: 613431083, Model: PonyRealism_v1.1_VAE, Version: 1.7.0, Sampler: Heun, Template: [object Object], CFG scale: 6, Clip skip: 2, Model hash: 5c20246ff3, ADetailer model: face_yolov8n.pt, Wildcard prompt: [object Object], ADetailer version: 24.1.2, Negative Template: [object Object], ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True