Scene 8: Characters
Celebrity Photos
This is a new feature of V5. Midjourney has met many celebrities in the V5 version. When describing the subject, you can add the name of a celebrity to generate their photo. We have generated several classic movie photos:
The first one is Keanu Reeves, the second one is Vito Corleone (I did not input the lead actor Marlon Brando), and both sets of prompts for the two pictures are very simple. You can try writing them yourself 😁
The answer is provided at the end of this chapter.
Clothing Model
Friends who work in clothing e-commerce may need this function more:
- Use AI to generate a virtual person
- Then upload a photo of goods (such as a piece of clothing) to AI
- AI blends the virtual person with the photo of goods together
To be frank, I think Midjourney should be able to do it, but currently its effect is not very good. To achieve this requirement, we only need to use the blend function mentioned in technique nine earlier.
The detailed method will not be repeated here. In general, I blended the product image and character image using blend function.
To verify consistency, I generated several Reeves' graphs and tried them separately, and the final output looked like this
Overall, it seems okay, but using it in product description images may not be very possible because there is still a significant difference between the clothes inside and the original image (you can see this by looking closely at the buttons).
However, I think Midjourney may optimize this feature in the future. Additionally, if users are able to train their own models, implementing this feature should not be difficult.
Old Photos
This is a branch of celebrity photo play. Recently, many similar photos have been seen. In fact, the prompt is very simple - just use translation software for both subject and background.
This is a branch of celebrity photo play. Recently, you may have seen many similar photos. In fact, the prompt is very simple.
| prompt | explain | |
|---|---|---|
| Type | photography | / |
| Body | a group of Chinese people gathered around Darth Vader | / |
| Background | on the street | / |
| Composition | null | Not important, just let AI play freely. |
| Lens | fuji film | In order to give the photos more flavor, a Fujifilm lens was added. |
| Style | style of 1990s | / |
Two sets of images were generated with a similar prompt:
Tip 14: Adding Style - Year
In the old photo scene above, we used a 1990s style. Strictly speaking, this 1990s style is more suitable for the subject or environment rather than being placed in the style category.
This is because not only do we want the subject to be from the 1990s, but also want the generated image style to be from that era.
This is the last usage of stylistic words: time period.
By modifying two prompts from the illustration chapter and changing them to "style of 1920s" instead of including an artist's name, we can see that both rose and cat illustrations have a touch of that era's flavor.
The picture in the upper right corner of the cat illustration is particularly amusing.
Answer to the celebrity photo:
For Reeves, it's very simple. Just describe the subject and then indicate that he is wearing iconic clothing while walking in heavy rain.
Keanu Reeves, wearing a black long leather coat, walking down the street in the rain
Then the Godfather's one is even simpler:
Vito Corleone sitting on a chair
Tip 15: How to make the people generated by Midjourney more diverse?
Here, diversity refers to differences in height, weight, and even some physical characteristics.
You may have noticed that most images generated by AI feature very attractive and handsome characters.
However, this is not necessarily a good thing as it reflects societal biases to some extent.
Therefore, in this chapter I will focus on how to generate more diverse human figures using Midjourney.
- Method 1: Slider Method
In the previous chapter, I introduced a method for increasing weights which can also help you generate characters with different body types. Below is an example from the official Midjourney community:
Plus-sized ginger female dressed in bohemian vest, maxi skirt, espadrilles, walking across autumn campus courtyard, morning atmosphere:: plus-sized::-0.3 --ar 2:3
You can control the parameters behind 'plus-sized' to make the characters in the picture fatter or thinner. It is important to note that a negative number does not necessarily mean thinner, but rather weakens the weight of 'plus-sized'.
If your prompt word is 'thin' and you add a negative slider parameter, then it will make them fatter instead.
- Method 2: Direct Description.
Midjourney knows some diversity-related words, such as:
blind,
deaf,
cerebral palsy,
accessibility
