AI 作画《Concept Art概念艺术》| 用stable diffusion生成

前言

  • “Concept Art”是一个艺术门类,即所谓“概念艺术”、“概念设计”,也称为“初步设计”。一般常见于影视或游戏设计中,服务产品最初的视觉效果,是游戏或影片的核心工作内容之一。

  • 下面让我们使用当前很火的扩散模型stable diffusion,生成一些概念艺术作品!看看AI作画的能力如何!

1、古代印度尼西亚村民,电影,详细,大气,史诗,概念艺术

  • 输入:ancient indonesia, indonesian villagers, punakawan warriors and priests, cinematic, detailed, atmospheric, epic, concept art, wimmelbilder, matte painting, background mountains, shafts of lighting, mist,, photo – realistic, concept art,, volumetric light, cinematic epic + rule of thirds | 3 5 mm, 8 k, corona render, movie concept art, octane render, cinematic, trending on artstation, movie concept art, cinematic composition, ultra – detailed, realistic, hyper – realistic, volumetric lighting, 8 k

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第1张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第2张图片

2、寺庙, 森林, 楼梯, 柱子, 电影, 详细, 大气, 史诗, 概念艺术, 哑光绘画, 背景, 薄雾, 真实感, 概念艺术

  • 输入:temple in ruines, forest, stairs, columns, cinematic, detailed, atmospheric, epic, concept art, Matte painting, background, mist, photo-realistic, concept art, volumetric light, cinematic epic + rule of thirds octane render, 8k, corona render, movie concept art, octane render, cinematic, trending on artstation, movie concept art, cinematic composition , ultra-detailed, realistic , hyper-realistic , volumetric lighting, 8k –ar 2:3 –test –uplight

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第3张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第4张图片

3、城市,现实主义,电影,现实主义,概念艺术

  • 输入:city made out of glass : : close shot : : 3 5 mm, realism, octane render, 8 k, exploration, cinematic, trending on artstation, realistic, 3 5 mm camera, unreal engine, hyper detailed, photo – realistic maximum detail, volumetric light, moody cinematic epic concept art, realistic matte painting, hyper photorealistic, concept art, volumetric light, cinematic epic, octane render, 8 k, corona render, movie concept art, octane render, 8 k, corona render, cinematic, trending on artstation, movie concept art, cinematic composition, ultra – detailed, realistic, hyper – realistic, volumetric lighting, 8 k

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第5张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第6张图片

4、概念艺术,亚光绘画,指环王,权力的游戏,雾,现实主义,概念艺术,电影史诗

  • 输入:forest wanderer by dominic mayer, anthony jones, Loish, painterly style by Gerald parel, craig mullins, marc simonetti, mike mignola, flat colors illustration, bright and colorful, high contrast, Mythology, cinematic, detailed, atmospheric, epic , concept art, Matte painting, Lord of the rings, Game of Thrones, shafts of lighting, mist, , photorealistic, concept art, volumetric light, cinematic epic + rule of thirds | 35mm| octane render, 8k, corona render, movie concept art, octane render, 8k, corona render, cinematic, trending on artstation, movie concept art, cinematic composition , ultra detailed, realistic , hiperealistic , volumetric lighting , 8k –ar 3:1 –test –uplight

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第7张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第8张图片

5、哥特式风格城堡,电影,概念艺术,雾

  • 输入:Environment castle nathria in world of warcraft ::gothic style fully developed castle :cinematic, raining, night time, detailed, epic , concept art, Matte painting, shafts of lighting, mist, photorealistic, concept art, volumetric light, cinematic epic + rule of thirds, movie concept art, 8k, cinematic, trending on artstation, movie concept art, cinematic composition , ultra detailed, realistic , hyper realistic , volumetric lighting , 8k –ar 3:1

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第9张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第10张图片

6、帐篷未来主义家庭小屋,模块化,绘画概念艺术

  • 输入:cabela’s tent futuristic pop up family pod, cabin, modular, person in foreground, mountainous forested wilderness open fields, beautiful views, painterly concept art, joanna gaines, environmental concept art, farmhouse, magnolia, concept art illustration by ross tran, by james gurney, by craig mullins, by greg rutkowski trending on artstation

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第11张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第12张图片

7、一只可爱的魔法飞行狗,由迪士尼概念艺术

  • 输入:a cute magical flying dog, fantasy art drawn by disney concept artists, golden colour, high quality, highly detailed, elegant, sharp focus, concept art, character concepts, digital painting, mystery, adventure

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第13张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第14张图片

8、蜘蛛侠和蝙蝠侠之间的、混合超级英雄,概念艺术

  • 输入:clear portrait of a superhero concept between spiderman and batman, cottagecore!!, background hyper detailed, character concept, full body, dynamic pose, intricate, highly detailed, digital painting, artstation, concept art, smooth, sharp focus, illustration, art by artgerm and greg rutkowski and alphonse mucha

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第15张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第16张图片

9、数字概念艺术,唐朝服饰

  • 输入:a digital concept ar by artgerm and greg rutkowski and alphonse mucha. clear portrait of a lonely attractive men in uniform of tang dynasty!! heavy armored cavalry of the tang dynasty!! light effect. hyper detailed, character concept, full body!! dynamic pose, glowing lights!! intricate, elegant, artstation, concept art, smooth, sharp focus, illustration

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第17张图片 AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第18张图片

注:以上输入模型的文本描述参考于:https://mpost.io/best-100-stable-diffusion-prompts-the-most-beautiful-ai-text-to-image-prompts/

扩散模型原理

  • 扩散模型是一种概率模型,通过逐步去噪一个正态分布变量来学习数据分布p(x),对应于学习长度为t的固定马尔可夫链的反向过程。模型可以通过训练去噪自编码器来实现(T = 1…T),它们被训练来预测其输入的去噪变体,而是输入的噪声版本,其训练函数:

  • 扩散模型原理方面,包括:前向的加噪和逆向的去噪;

  • 前向的加噪:给定初始数据(比如图像),一步一步加噪得到、、···、、、···、,最后的完全是一个正太分布噪音数据。其中,每一步的加噪,不妨记为 分布表示;对应地,实际的去噪分布记为 ;

  • 逆向的去噪(生成):一步一步去噪,得到、···、、、···、、,不妨记为 分布表示;

  • 事实上,扩散模型的训练就是用网络(待学习参数为)预测所加的噪音变量分布。用KL散度衡量实际的去噪分布 和由网络学习的去噪分布:即。经一系列推导化简、等价为 。这里的是高斯噪声,预测网络以含噪图片作为输入,预测所添加的噪声。

  • 即可以理解为,扩散模型的训练目标是,希望预测的噪声和真实噪声一致

猜您喜欢:

深入浅出stable diffusion:AI作画技术背后的潜在扩散模型论文解读

7b7f0d7afe2c880962ed1e38b3d64332.png 戳我,查看GAN的系列专辑~!

一顿午饭外卖,成为CV视觉的前沿弄潮儿!

最新最全100篇汇总!生成扩散模型Diffusion Models

ECCV2022 | 生成对抗网络GAN部分论文汇总

CVPR 2022 | 25+方向、最新50篇GAN论文

 ICCV 2021 | 35个主题GAN论文汇总

超110篇!CVPR 2021最全GAN论文梳理

超100篇!CVPR 2020最全GAN论文梳理

拆解组新的GAN:解耦表征MixNMatch

StarGAN第2版:多域多样性图像生成

附下载 | 《可解释的机器学习》中文版

附下载 |《TensorFlow 2.0 深度学习算法实战》

附下载 |《计算机视觉中的数学方法》分享

《基于深度学习的表面缺陷检测方法综述》

《零样本图像分类综述: 十年进展》

《基于深度神经网络的少样本学习综述》

AI 作画《Concept Art概念艺术》| 用stable diffusion生成_第19张图片

你可能感兴趣的:(机器学习,深度学习,人工智能,计算机视觉,tensorflow)