开发语言：WPF C#

大模型：majicmixRealistic_v7

PC离线搭建：离线运行模式，本机运行无需联网。

麦橘写实 V7是由国内知名 AI  模型师 "麦橘" 开发的超写实通用模型，以人物摄影级真实感和皮肤质感表现著称，在 CivitAI 平台拥有4.93 分 (5 分制)高评分和70 万 + 下载量，是当前最受欢迎的写实类模型之一。

核心优势：

超高人像真实度，能完美呈现五官细节、发丝和皮肤质感
强大光影表现，擅长处理明暗对比和立体感
广泛场景适应性，适用于人像、风景、产品等多种创作
对 LoRA 和插件兼容性极佳，扩展空间大。

由于模型转换onnx可能有些鸡肋，CPU大图带不动，GPU小尺寸无压力(大尺寸吃力)，但总算成功了。提示词：支持中文 但倾向英文.

更新：2026年4月11日

1、同时生成多张图片（同时生成4张仅用1分30秒左右，显卡：GEFORCE 3060Ti 下测试结果）；

2、添加外置英语翻译字典；

3、修改BUG。

更新：2026年4月10日

1、提示词添加；

2、添加图生图；

3、修复BUG。

AI 强效负面：部分代码

private void InitializeNegativePresets()
{
    negativePresets["通用质量增强"] = @"lowres, error, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, text, logo, ugly, duplicate, morbid, mutilated, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, bad anatomy, bad proportions, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, bad feet, cross-eyed, missing fingers, bad hands";

    negativePresets["人物写实专用"] = @"ugly, deformed, disfigured, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, disconnected limbs, mutation, ugly face, ugly eyes, cross-eyed, asymmetric eyes, cloned face, big nose, long neck, malformed hands, extra fingers, missing fingers, fused fingers, poorly drawn hands, text, logo, watermark, blurry, out of frame";

    negativePresets["二次元/动漫人物"] = @"lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, deformed, disfigured, ugly, cross-eyed, extra arms, extra legs, missing arms, missing legs, mutated hands, fused fingers";

    negativePresets["高质量人像(去瑕疵)"] = @"acne, spots, freckles, skin blemishes, mole, wrinkles, dark circles under eyes, fat, ugly, deformed, bad anatomy, cross-eyed, uneven eyes, bad teeth, gross proportions, malformed limbs, extra fingers, missing fingers, blurry, text, watermark";

    negativePresets["风景纯净专用"] = @"buildings, city, urban, people, humans, person, man, woman, child, cars, vehicles, roads, trash, litter, pollution, disfigured, bad anatomy, extra limbs, mutated, deformed, blurry, lowres, worst quality, text, watermark, signature, logo";

    negativePresets["建筑/室内专用"] = @"ugly, deformed, disfigured, humans, people, person, blurry, lowres, bad anatomy, extra limbs, text, logo, watermark, signature, crooked walls, messy, dirty, bad perspective";

    negativePresets["产品摄影专用"] = @"people, humans, hands, fingers, text, logo, watermark, signature, blurry, lowres, worst quality, ugly, deformed, disfigured, extra objects, messy background, dirty, smudges, scratches";

    negativePresets["极简/干净背景"] = @"text, logo, watermark, signature, people, humans, animals, extra objects, debris, messy, dirty, blurry, low quality, worst quality, cropped, out of frame";

    // ✅【修复】删除了不兼容的特殊嵌入词 ng_deepnegative_v1 等，99%模型兼容
    negativePresets["AI强效负面(推荐)"] = @"(worst quality:2), (low quality:2), (normal quality:2), lowres, watermark, monochrome, skin spots, acnes, skin blemishes, age spot, extra fingers, fewer fingers, bad hand, bad anatomy, text, error, missing fingers, cropped, jpeg artifacts, signature, blurry";

    // ✅【修复】删除了会崩溃的畸形关键词
    negativePresets["去AI畸形/扭曲专用"] = @"bad anatomy, bad hands, missing fingers, extra fingers, deformed, fused fingers, too many fingers, extra arms, extra legs, malformed limbs, long neck, cross-eyed, ugly, blurry, out of frame";

    negativePresets["自定义"] = "";
}

 视频大模型 需要下载的模型文件（Wan2.2 14B）

文件	说明
Wan2.2-T2V-A14B-LowNoise-Q8_0.gguf	低噪声扩散模型
Wan2.2-T2V-A14B-HighNoise-Q8_0.gguf	高噪声扩散模型
umt5-xxl-encoder-Q8_0.gguf	T5XXL 文本编码器
vae.safetensors	VAE 模型

大模型更加贴合英文所以：

["古建筑"] = "ancient architecture", ["名胜古迹"] = "scenic spots and historical sites", ["世界奇观"] = "world wonder", ["春天"] = "spring", ["夏天"] = "summer", ["秋天"] = "autumn", ["冬天"] = "winter", ["东"] = "east", ["西"] = "west", ["南"] = "south", ["北"] = "north", ["日出"] = "sunrise", ["日落"] = "sunset", ["山川"] = "mountains and rivers", ["河流"] = "river", ["湖泊"] = "lake", ["森林"] = "forest", ["草原"] = "grassland", ["沙漠"] = "desert", ["海洋"] = "ocean", ["海岛"] = "island", ["雪山"] = "snow mountain", ["峡谷"] = "canyon", ["瀑布"] = "waterfall", ["城堡"] = "castle", ["宫殿"] = "palace", ["寺庙"] = "temple", ["园林"] = "garden", ["小镇"] = "small town", ["都市"] = "city", ["夜景"] = "night view", ["云海"] = "sea of clouds", ["星空"] = "starry sky"...

问豆包：豆包，你的生图模块，提示词是不是用户输入后，你根据关键字补充全了才生成的图？？？？

是的，你理解得基本没错，可以直白跟你说清楚：

你输入的提示词，我不会直接原封不动丢给画图模型，而是会：

1. 先解析你写的关键词、风格、构图、画质要求
2. 自动帮你补充、润色、扩展成一段更完整、更符合模型习惯的专业提示词
3. 同时自动加上画质、分辨率、采样参数等
4. 再用这段完整提示词去生成图片

简单说：

• 你写的 = 核心意图
• 我处理后 = 最终真正用来画图的完整提示词

所以经常会出现：你写得很简单，但出来的图细节很丰富，就是因为中间我帮你补全优化了。

直接给你看真实对比示例，一眼就懂。

---

示例：你输入的提示词

一只可爱的白色小猫，在草地上睡觉

---

我实际传给画图模型的完整提示词

A cute white kitten, sleeping peacefully on soft green grass, warm sunlight, fluffy fur, delicate details, cinematic lighting, high detail, 8k, ultra-detailed, masterpiece, best quality, shallow depth of field, vibrant colors, natural environment

同时还会自动带上：

• 分辨率：1024×1024
• 采样步数、画质增强、负面提示词（避免模糊、畸形、低质）

---

简单总结

你只说：小猫在草地睡觉我帮你补成：什么样的小猫 + 光线 + 画质 + 构图 + 细节 + 专业参数

所以你写得越简单，我补得越多；你写得越专业，我就越少改动，更尊重你的原意。