How AI helps to create better images

Midjourney is an artificial intelligence (AI) programme capable of creating images using AI. The programme originates from the Midjourney Research Institute, a renowned research institute based in San Francisco, California, USA. Midjourney has been available in open beta form since July 2022. 

The artificial intelligence is controlled via commands to a Discord bot. In order for the AI Midjourney to create its AI art, users must enter a prompt. This prompt then serves as the basis for the system to create the image. It is important to note that the command "/imagine" must always appear at the beginning of the prompt. This is followed by the actual description of the image - or rather: the content that should be seen. With Midjourney, it is sufficient to give a brief description of what the picture should contain. 

The AI Midjourney displays the images in a preview. The system gives a total of four possible drafts. Users can then decide for themselves which design appeals to them most. Now users still have the option of scaling the images (U1-U4) or getting different variants of the image (V1-V4). If there is no image that matches the prompt or contains errors, users can have completely new images generated - until they are satisfied with the result. The AI is able to generate a variety of images. Whether product pictures, portraits of people, landscape shots, comic pictures or animal pictures - Midjourney generates exactly the picture you need.

Proper prompting: how to get the best results
Prompting is an important part of the interaction process with AI systems like Midjourney. It refers to the text or instructions you give to the AI to guide it on what to do. We've put together some helpful tips to help you create even more effective prompts for Midjourney:

Be specific:
The more detailed and specific your description is, the better Midjourney will be able to meet your needs. For example, if you want to create an image of a forest at sunset, it would be helpful to describe details such as the types of trees, the colours of the sky and other aspects of the scene. However, be careful not to make the prompt longer than necessary. Short concise prompts - regardless of grammar or spelling - are enough for the AI to generate results.

Use metaphors and similes:
If it's difficult to describe exactly what you want, you can use metaphors or similes to give your description more context. For example, if you want to create an image of a person who looks "as strong as a lion", Midjourney would try to create an image that represents these qualities.

Be willing to experiment:
Different formulations can lead to different results. So if you don't get exactly what you imagined, try changing your description and try again. You will see: Practice makes perfect.

Give feedback:
Midjourney offers the possibility to choose between four generated drafts and edit them or generate new ones if they don't meet your expectations. Use these features to improve your results and help Midjourney better understand your preferences.

Understand the limitations of AI:
Although AI systems like Midjourney are incredibly powerful, there are still limits to what they can do. For example, it can be difficult to generate very abstract concepts or very complex scenes. It's also important to note that AI doesn't "think" or "understand" in the same way as a human, so its interpretations of prompts can sometimes be unexpected. Misunderstandings are also not uncommon when you use terms that have a double meaning. For example, the word castle could be understood as a fairytale castle or a padlock. So think carefully about how you formulate your prompts and which terms you use.

Learn from examples:
Look at the examples of other users to see what types of prompts lead to successful results. Learn from your own mistakes and those of others. This way you can perfect your prompts and get the best results.

Add moods and emotions:
Emotional context can help the AI create an image that captures the desired atmosphere. For example, you could add terms like "peaceful", "happy" or "gloomy" to your description to define the mood of the image you want.

Use the story:
If you want to create an image that represents a particular story or concept, it can be helpful to include that element in your prompt. For example, a prompt like "A knight preparing for a dangerous quest" might prompt AI Midjourney to create an image that includes these narrative elements.

Be patient and open:
The AI-generated artwork may not always match your original vision exactly, but can still be interesting and valuable. Be open to surprising or unexpected results and see them as part of the creative process. After all, Midjourney can only deliver as good results as the prompts you give to the AI. Be clear in advance about what you want to have - but with some room for interpretation. Maybe the AI will even give you new ideas.

Ask for several drafts:
Midjourney first creates four possible drafts based on your prompt. If none of them matches your ideas, you have the option to discard all four drafts and request four completely different images. So don't be afraid to start again. Try to optimise your prompt further and see exactly which content Midjourney had problems with when creating it.

Midjourney is a good tool to generate visually appealing images quickly and easily. But not every prompt works right away. That's why a certain amount of experimentation is required, especially on the user side. In our test, we also made sure to give Midjourney just enough input so that it is not overwhelmed and can limit itself to the essential content.