Midjourney绘画精准控制画面元素的技巧(数字篇)

使用Midjourney生成图像时,总是抱怨无法控制画面中元素的比例、结构、大小、轻重、数量等,更多的同学把这些归结于Midjourney的可控性弱。其实,如果加入更加具体的指标,这些问题就会豁然得到解决,而这些技巧往往被忽略了。

▶ prompt中加入身高数据
我想生成一个可爱的小公主和魁梧的国王,最开始我们用普通的prompt进行了描述,如下:
prompt

full body shot, side view, a princess looking at a king who stands in front of him, in a Disney-style, with natural lighting, ultra-high precision, high detail, 8K resolution, created using C4D and Blender, and rendered using OC rendering. --v 5.2

完全没有表现出公主的娇小可爱和国王的高大慈祥
我们只需要在prompt中加入身高数据,其它基本都没变,具体的描述词如下:
prompt

full body shot, side view, a 1.2-meter-tall princess looking at a king who stands 1.85 meters in front of him, in a Disney-style, with natural lighting, ultra-high precision, high detail, 8K resolution, created using C4D and Blender, and rendered using OC rendering. --v 5.2

这次的生成结果就令我们非常满意了,如下图
图片
娇小可爱的小公主和高大魁梧慈祥的国王形成了鲜明的对比
图片
图片
调整前后效果对比

再比如,我们要突出武士面对身躯庞大的巨龙时的勇气,就需要突出巨龙和武士之间巨大的个体差距,这时候我们可以加入巨龙的高度(6m)和武士的高度(1.8M),这样的画面视觉冲击力就强的多,完全把巨龙的庞大、恐怖、威压、力量体现的淋漓尽致,更能体现出武士面对这样的庞然大物不退缩、不胆寒,勇敢面对,豪气冲天的形象。
prompt

Mythical scene, a 6-meter-tall giant dragon facing the camera, with a 1.8-meter warrior facing the giant dragon, in a symmetrical composition, ultra-high precision, high detail, 8K resolution, created using C4D and Blender, and rendered using OC rendering --v 5.1

图片

一人一枪单挑恶龙

当然,生成个体时同样可以加入数据,比如以下的例子,身高1.67M体重54KG,是不是模特好身材的最大保障?图片
prompt

A real photo of a woman with straight reddish hair. The woman is elegantly dressed, with a beautiful dress, she is 1.67 meters tall and weighs 54 kg. A photo with an urban background, and also see her legs

图片

身高和体重什么比例身材最好?

▶ Prompt中加入尺寸数据
尤其是在建筑效果图或者展会效果图等方面,加入具体的尺寸或者面积,生成的结果更接近于现实。如下图就是明确指定了尺寸是3m×3m,高度是2.4m,最终生成的效果完全模拟了展会现场。
prompt

The prompt features a 3x3-meter stand with a height of 2.40 meters for a company selling shrimp in the United States. The stand has an attractive and eye-catching design that highlights the freshness and quality of the shrimp offered by the company. The stand is decorated with blue and white colors, representing the freshness of the sea and the purity of the shrimp. On the front of the stand, the name and logo of the company are displayed, standing out with LED lights to attract customer attention. Inside the stand, there is a stylish and well-lit counter, where customers can see and select the fresh shrimp available. There is also a tasting section for customers to try the delicious variety of products offered by the company. The stand features images and graphics showing the quality and production process of the shrimp, generating customer confidence in the freshness and origin of the products. The stand's staff is uniformed and trained to provide excellent customer service, answering questions and recommending different purchase options. This stand is a perfect display to promote and sell fresh, high-quality shrimp, highlighting the company's dedication to offering top-notch seafood products to customers in the United States.

图片

生成平面效果图
直接标明宽20米,高3米,5平方米的厨房,10平方米的会议室
prompt

Architectural office in a modern design where 4 groups of 2 people can work in a rectangular shape, 20 meters wide, 3 meters high, 250 cm high, 3 meters front entrance, 5 square meters toilet and 5 square meters kitchen, 10 square meters meeting room at the back.

图片

图片

▶ prompt中加入年龄数据
在人物描述时,很多同学都只是简单的描述为女孩,女人,老人等,往往生成的图像和你心目中的形象相差甚远,我们可以试着加入她们的年龄,更能体现出相应年龄段的人物特征。我在一个提示词中就加入了3个年龄,从中我们也可以看出加入年龄数据之后生成的图像更贴合实际,甚至可以体现出同一个人在不同的年龄段呈现出的气质和特征,在以下的promot我们就加入了18岁、30岁、70岁这样的年龄数据,是不是比单纯的描述“三个女人”来的更精确?
prompt

Realistic portraits, facing the camera, of an 18-year-old girl, a 30-year-old woman, and a 70-year-old woman, with high precision, 8K resolution, and high detail. --ar 16:9 --v 5.2 --style raw

图片
18岁-30岁-70岁
▶ prompt中加入个体数量
prompt

heartwarming scene, 5 adorable children playing with building blocks inside a toy house, light, image is in 8K

图片

其实,AI绘画固然能够提高工作效率和质量,但真正要掌握AI绘画的技巧,也需要点点滴滴的积累,今天给大家分享只是一个小技巧,想要获得更多AI绘画甚至AI自然语言大模型的资料,大家可以在公众号后台回复【知识库】获取

文 | 老魏职道

编辑 | Robin

THE END