I deleted and rewrote the Text Prompt part three times, but I was not satisfied with it every time. In the first two times, I hoped to teach everyone how to write the Text Prompt of Midjourney through one or several universal templates, but after trying many times, I found that:

1. The amount of information in a picture is very large. The universal template does allow you to generate a passing picture, but it is difficult for you to generate a completely satisfactory picture.
2. To generate a high-score image, not only prompt engineering is needed, but also some aesthetic knowledge. The parameters used in different scenes are different, so sometimes using templates is a bit rigid.

Therefore, I don’t want to teach you the so-called universal template directly, but teach you how to write a good text prompt through actual scenarios, because the information you tell the model is different in different scenarios.

And I think that knowing why is far more important than knowing how to do it. When you know why, when you encounter a new scenario, you can also deconstruct what is needed in the prompt well, instead of blindly using a template.

Finally, there are many picture scenes, I will try to share some that you can use in your daily work. I hope that AI can improve your work efficiency instead of being a disposable toy.

Text Prompt Let me start with an introduction Midjourney V5 has a significant improvement Stock Photo scene.

What is a Stock Photo?

Stock Photo literally means stock photo. You can usually find it on some stock photo websites. These pictures usually come from some photographers or designers. Some pictures have copyright reasons, so you need to pay if you want to use them.

Most users of Stock Photo are design companies or advertising companies. You should often see this kind of pictures, such as the most classic photo of two people shaking hands:

I think AI-generated images will have a huge impact on the photo library, and the V5 version basically meets my Stock Photo needs.

Technique 1: Copy

I think learning picture-based prompts is similar to learning to draw, and the best way to learn is not to use templates directly.

Instead, you can copy a real picture or a picture generated by someone else. If you are not good at English, you can also write it in Chinese first and then let ChatGPT translate it. After you copy a few pictures, you will slowly understand how to make similar pictures.

Take the handshake picture above as an example. Let's take a closer look at the picture above. What are the elements in the picture?

1. The first subject is two hands, which are then clasped together, and appear to be two Asian men.
2. The second two men were both wearing suits.
3. The third background looks like the gate of an office building, and the two people may be shaking hands and saying goodbye. And the background is deliberately blurred, or it was taken with a camera.

To summarize, the general information is:

• Subject: Two Asian men in suits shaking hands and saying goodbye
• Scene: Office building entrance
• Image style: stock photo, taken with a camera,

At this point, we can try to write a prompt (if you feel your English level is not good enough, you can also try to translate it using translation software).

stock photo of two Asian men in suits shaking hands,say goodbye in front of the main entrance of the office building,taken with Canon

The results generated by Midjourney are as follows:

Emm 🤔 It seems to be different from what we expected. Don't panic, you will definitely encounter this kind of problem when you first use Midjourney. The important thing is to try more.

Let's analyze why Midjourney generates such an image.

1. First of all, the main subject of the picture, or the focus, we only need a "handshake" instead of two people.
2. The photo mode doesn't seem to be able to achieve background blur? It's more like an image style. Picture 4 looks like an old photo style.

Then let's adjust the prompt and add focus and background blur keywords:

stock photo of two Asian men in suits shaking hands,say goodbye in front of the main entrance of the office building, focus on two hands, taken with Canon, background bokeh

Let's take a look at the generated results, which are much better:

Figures 1 and 4 basically meet our needs. Figures 2 and 3 can also meet our needs after some cropping. However, it should be noted that Midjourney still has some problems in the generation of human hands. If you look closely at Figures 2 and 4, one of the people has 6 fingers on his hand 😂, but I think it should be fixed in the future.

Let's summarize the prompt, which is divided into the following parts:

• Part 1 (red line): Describe the main body of the content you want.
• Part 2 (blue line): describes the subject background/environment.
• Part 3 (yellow line): The focal point of the photo.
• Part 4 (green line): Photo style or special requirements.

After summarizing, have you come up with a template? 😁

Tip 2: Experiment

In the above case, I would also like to teach you a second trick: experiment more.

If the image generation does not meet your expectations, don’t panic, analyze the problem, and then use the control variable method to adjust the images one by one. Don’t be anxious. There is another place in the prompt I wrote above. I don’t know if you have noticed it. It is the stock photo at the beginning. What will happen if you delete these two words?

two Asian men in suits shaking hands,say goodbye in front of the main entrance of the office building, focus on two hands, taken with Canon, background bokeh

The generated result still meets the requirements, and the number of fingers is also accurate, which shows that this stock image does not have a big impact on the model.

Tip 3: Make good use of Image2Image function

In the Stock Photo scene, there is another very powerful technique. I didn’t think it was suitable to teach this method at first because it would have too much impact on the Stock Photo library.

But based on the principle of tool neutrality, and this technique can actually be used in many scenarios (such as generating avatars, etc.), so I still feel it is necessary to teach you.

You will generally encounter the following problems when using the Stock Photo library:

1. The image is copyrighted and cannot be used commercially, or a fee is required.
2. Some pictures have been used by many people, and other people can tell at a glance that they are stock photos.
3. The content of the picture generally meets the requirements, but the details do not meet the requirements. For example, the picture above shows two Asian men shaking hands. It might be better to replace it with a man and a woman, and one of them is black.

The best way to solve these three problems is to let AI modify the original image. To let AI modify the image, you only need to use the Image2Image (or Blend) function. The steps are as follows:

1. Send any good stock photos you see to Midjourney Bot (I’ll use the handshake photo above as an example)
2. Right click to copy the link of the image, and then paste it into the input box
3. Add a space after the link
4. Then enter the content you want, such as changing one person's hand to a black person and one person to a woman:

one African-American hand and one Asian woman hand

The generated result is this. I didn’t mention any suits or scene background information in the prompt. I just said I wanted an African American hand and an Asian woman’s hand:

Putting aside the issue of 6 fingers, isn’t this technique very efficient? However, please note that the blend function (which will be discussed in the subsequent tutorial) is more suitable for merging two pictures, and picture + text is not very stable. Please be patient and test it more.