Artificial intelligence has reached an incredible level! It is now possible to generate beautiful images in a few seconds thanks to text-to-image AI like Midjourney. To use it, you simply write a Prompt that allows the Midjourney bot to understand the characteristics of the image you ask it to produce.
This article gives you advanced techniques to create Prompts that provide better results, no matter how good you are with AI.
Text-to-Image AI: What is an Image Generation Prompt?
The term that comes up most frequently in tools using AI is Prompt. Simply put, a prompt is a small text that contains the description of the task that the artificial intelligence must perform. Text-to-image AIs can generate images from text or from an image with text. You are going to discover the different types of prompt possible for the generation of images.
Prompt with text
The basic Prompt to send to the bot to create an image is a small text that clearly describes what you want to obtain as a result. To get a better result you should give the name of the object and add keywords so that the AI understands what it is asked to do. The more detail you provide, the better the result will be.
An example of a text Prompt is :
Prompt: The Earth in the solar system, dark background
Prompt with an image
You can also make prompts using an image as a basis for generating a new image. This influences the style and colors of the final result. The bot uses this image to get the precise characteristics to give to the final result.
Prompt with an image and text
The Prompt can be made up only of an image or an image with text. This allows you to get a result closer to what you want, because the image is a model for the AI. The Prompt image provides the colors and other visual characteristics while the text provides some additional details about the expected result.
Rules for writing a good Prompt
To create an effective Prompt that delivers good results, it is important to follow some basic rules. While there is no one-size-fits-all method, here are some tips to help you write a Prompt that is tailored to your AI needs.
1. Use between 3 and 15 words
It is best to write Prompts that contain between 3 and 15 words. This keeps the request specific enough without being too long and complex.
2. Specify the subject and use descriptive adjectives
Specifying the subject, whether it’s a person, object or place, is essential to getting relevant results. Also use adjectives that describe the subject, such as “broken flower pot”.
3. Avoid abstract subjects
Abstract topics such as love, joy, or hate can make it harder for artificial intelligence to understand the request. Therefore, it is best to avoid them.
4. Describe in as much detail as possible
The more details you provide in the Prompt, the better the chances of getting an accurate result that meets your expectations. So don’t hesitate to describe what you want as precisely as possible.
5. Use keywords to indicate the type of image you want
If you are looking for a certain type of image, consider including keywords that indicate this. For example, “photo”, “3D sculpture”, “logo” or “illustration”.
By following these tips, you will increase your chances of getting satisfactory results when using Midjourney.
MidJourney: become the master of Prompts
Midjourney is an innovative tool based on artificial intelligence, allowing you to generate quality images from simple text. Whether you want to create a work of art, a portrait or a logo for your company, Midjourney offers fast and surprising results. To use this tool, you will first have to join the Midjourney Discord server.
- Join the Midjourney Discord server to access the tool.
- Once the server is joined, go and interact with the Midjourney bot in private message (Midjourney Bot#9282)
- Write a text command (the Prompt) describing the image you want to obtain by starting with the command /imagine to initiate the bot
- Wait a few seconds for the artificial intelligence to generate your image.
- Admire the result and adjust your prompt if necessary to get the perfect image.
Generate an image with text with Midjourney
Like most AI-based platforms, Midjourney uses Prompts to tell the bot what to do. The Prompt starts with the keyword /imagine followed by the description of the image you want to generate.
The following Prompt is an example of use and you can see the impressive result in the following image, and all this in just a few seconds.
/imagine Albert Einstein eating a steak, horse in the background
You can choose the style of the image among the many styles that the platform offers. You just have to add the corresponding keyword in your Prompt.
Here are some examples of styles you can use to customize your images:
- Thirtiesstyle (Thirties): this style allows you to give your images a retro and elegant look, inspired by the 1930s.
- Classicism: This style is inspired by classic works of art to create images with a refined, timeless look.
- Map Style: This style is ideal for creating images that represent maps or blueprints.
Classic Style
Thirties Style
Map Style
It is important to note that you can write your Prompt in French, but it is better to write them in English for better results. You should also keep in mind that Midjourney is not very good at adding text to images.
Adding parameters to Prompt
Parameters are elementary options that are added at the end of the Prompt to change the way the image is generated. A parameter starts with ” — “, for example:
/imagine Football player on fire –s
The seed parameter
Since the Midjourney bot generates the images in a completely random way, entering an identical Prompt generates a different result every time. If you want to have a similar result each time to see how the modification of your Prompt affects the generated image, add the parameter “–seed [Number]” at the end of your Prompt. This way the image produced will be a copy of the previous one using the same Prompt.
/imagine soccer player hitting a stone ball –seed 10
The chaos parameter
The Chaos parameter noted “–chaos” or “–c” influences the variation of the initial grids of the image. Higher values of chaos will produce more unusual and unexpected results while lower values of chaos will produce more reliable and repeatable results. The parameter can take a value between 0 and 100, the default value being 0.
For example, the following Prompts give the images below.
imagine/ watermelon owl hybrid
imagine/ watermelon owl hybrid –c 100
The stylize parameter
This parameter is denoted ” –stylize” or ” –s” and is used to influence the degree of styling applied to the image. Low values will produce images that match the Prompt, but are less artistic. While high stylization values will generate images that are very artistic but less close to the description of the Prompt. The parameter can take a value between 0 and 1000, the default value being 100.
Using the Remix mode
The Remix mode uses the general composition of the original image as a template for generating the new image. Remixing can help change the frame or lighting of an image.
To activate Remix Mode, use the /prefer remix command or use the /settings command and activate the “Remix Mode” button.
Starting Prompt
Prompt with Remix
Make a Multi Prompt
You have the possibility to ask the Midjourney bot to take into account two or more different concepts by using the separator “:”. You therefore have the possibility of assigning a relative importance to each part of the Prompt.
For example, the following two Prompts generate the two images below.
/imagine hot dog
/imagine hot :: dog
Prompts weights
You can add a number immediately after the colon to assign a relative importance to that part of the Prompt.
In the previous example, the Prompt ” /imagine hot :: dog” produces a dog that is hot while in the Prompt ” /imagine hot :: 2 dog”, the word hot is twice as important as the word dog, which gives the image of a very hot dog!
/imagine hot :: dog
/imagine hot :: 2 dog
Make a combination of Prompts
You can combine several prompts into one. To do so, you just need to add a list of options in your Prompt separated by commas between braces {}.
/imagine a green bird
/imagine a red bird
/imagine a yellow bird
For example, the 3 previous Prompts can be combined to give the Prompt :
/imagine a bird {green, red, yellow}
For example, the following Prompt gives the 4 results below.
/imagine a naturalist illustration of a {pineapple, blueberry, rambutan, banana} bird
How to make a Prompt from an image on MidJourney
With the Midjourney AI,you can upload one or more images that will serve as a reference to create a new image that has similar attributes to the source images. You will discover how to create an advanced Prompt using an image with text to improve the result, just read the following sections.
In this example, two images were used to create an Advanced Prompt: one image representing a statue and another representing flowers. The result is surprising and demonstrates the power of Midjourney AI to create innovative images.
Statue Flowers
Statue + Flowers
Make a Prompt with an image from the Web
To make a Prompt from an image, you have to paste the URL of the image and add a description text, as well as additional parameters to complete the Prompt. The supported formats are: png, gif and jpg.
The general structure of the image Prompt is :
/image Image1 Image2 Description -Parameters
It is important to note that :
- Images are placed at the beginning of the Prompt;
- The description corresponds to a text Prompt;
- A Prompt must contain either two images or an image followed by a text;
- The URL of an image must be a direct link to that image.
Making a Prompt with a personal image
To use a personal image in a Midjourney Prompt, upload it to the Discord server. To upload an image:
- Click on the “More” sign next to the Prompt editor;
- Select “Upload a file”;
- Select your image;
- Type /imagine in the editor;
- Drag the image file into the editor to add the image URL to the Prompt. You can also right click on the image, select “Copy Link”, and then paste into the Prompt.
Conclusion
You now know how to make advanced Prompt using only text or combining images and text.
When using Midjourney AI, keep in mind that the results you get depend on the details you provide. The more precise and informative you are, the more the generated image will be in accordance with your expectations. So, don ‘t hesitate to add as many keywords as possible in your prompts to guide the Midjourney bot to generate an image that is very close to what you want.
