8 Tips For Writing Effective Midjourney Prompts

In this newsletter, read about:

• 🕵️‍♀️ 8 Tips For Effective Prompts in Midjourney

• 🗞 News and Top Reads

• 📌 AI Art Tutorial: Photoshop has AI

• 🎨 Featured Artist: Troy Guénard

• 🖼 AI-Assisted Artwork of the Week

• 🤓 How to Get Started with Generative AI?

🕵️‍♀️ 8 Tips For Effective Prompts in Midjourney

Midjourney is an extraordinary tool. Sometimes, it will produce stunning images with just two random words. Other times, it will drive you crazy by failing to fulfill even the simplest requests.

Initially, I believed that finding images I liked on the Midjourney feed, examining the prompts, and adapting them to my needs would be a straightforward process. Unfortunately, I soon discovered that a visually pleasing image does not guarantee a good and replicable prompt. There are various reasons for this discrepancy: (1) some individuals simply luck out with unexpectedly fantastic images (which happened to me quite often!), (2) others overload their prompts with redundant details that ultimately have no impact on the final result, and (3) often, people simply receive something entirely different from what they initially requested.

While learning from other people's prompts can be beneficial, it is not sufficient to grasp the fundamental principles. So, today, I want to share a collection of general recommendations for crafting effective text prompts in Midjourney. These suggestions are based on my research of Midjourney documentation, insights from the #prompt-faqs channel on Midjourney Discord, and other valuable resources. Some of these tips are strong recommendations rooted in expert knowledge of how the AI models behind Midjourney operate, while others serve as optional techniques to stimulate creativity, should you require an extra boost 🙂 

Without further delay, let's delve into these recommendations and unlock the full potential of Midjourney!

1. Write short and specific prompts

Midjourney prompts have a maximum character limit, although not as strict and precise as Twitter's 280 characters. Once you exceed a certain number of words, the influence of subsequent words on the final result becomes negligible. While different versions of Midjourney may handle these limits differently, as a general guideline, anything written after approximately 40 words is likely to be ignored, after 60 words it becomes highly likely to be ignored, and after 80 words it will almost certainly be cut off.

Furthermore, the length of your prompt also affects the level of competition among different elements for priority. Put simply, when you have a prompt consisting of just three words, there are only eight possible combinations of those words that can be expressed in the generated image. However, when your prompt extends to 40 words, the number of permutations becomes seemingly uncountable. While Midjourney has developed the ability to recognize the most significant words and likely combinations, having more words in your prompt reduces the chances of obtaining the exact combination you desire. So, if you are looking for a particular outcome, keep your prompts concise and specific.

2. Start with the most important keywords

Earlier Midjourney versions have a very strong slope of decreasing influence:

• Words 1–5 are very influential and will likely show off from the very first roll.

• Words 5–20 are also influential, but you may need 2-3 rerolls to have them show off on the image.

• Words 21–40 are still in play but may require too many rerolls to show off.

• Words 40+ are very likely to appear ignored.

In the latest versions of Midjourney, the rule of word order remains relevant, but additional factors are now taken into consideration, such as token strength. Token strength refers to Midjourney's learned perception of the significance of each word or phrase in the prompt. While you cannot directly influence token strength, you do have control over the word order in your text prompt.

So, start your text prompt with the most important keywords, but also, …

3. Use grammatically correct sentences in prompts

Starting from v5, Midjourney has implemented natural language processing (NLP) technology. As a result, using grammatically correct, full-sentence prompts tends to yield more relevant and coherent results compared to fragmented or nonsensical phrases. This means that proper grammar partially compensates for the diminishing influence discussed earlier. In fact, writing in a structured manner, adhering to the grammar rules you learned in school, becomes more important than focusing solely on word order.

4. Follow the prompt template

This is not a strong rule, but following a certain prompt structure helps many people to improve their prompting. For example, when relevant, your prompt may include the following elements in the suggested order:

1. Type of image (e.g., a photo, an illustration, a painting, etc.)

2. The main object described with relevant characteristics (e.g., a mystical woman with flowing hair in an ethereal dress).

3. Other details and surroundings (e.g., in a misty forest, amidst a field of sunflowers).

4. Style details (e.g., fine art photography, in the style of Picasso, etc.).

5. Parameters (aspect ratio, level of stylization, model, etc.)

5. Use Sliders and Cowbells tricks

If you find that Midjourney seems to disregard certain parts of your text prompt, you can utilize text weights to adjust the influence of specific words.

You put :: at the end of the base prompt, which is equivalent to assigning it a weight of 1. Then, you want the weights of different prompt parts to approach 1, but not exceed that value so that they don’t become more powerful than your base prompt. The recommended positive values for text weights are 0.25, 0.3, 0.5, 0.6, and 0.7. The recommended negative values are -0.7, -0.6, -0.5, -0.3, and -0.25.

In the #prompt-faqs channel, they call it a “slider” method because you kind of use a virtual slider to decide how much weight you want to put on each part of your prompt.

When the slider method doesn’t work for you, try a very simple “cowbell” method. It basically means repeating multiple times the individual words that are being ignored.

6. Don't overload your prompts with "quality enhancement" keywords

When attempting to obtain photorealistic images, people often include words like "photorealistic," "realistic," "ultra-detailed," "8K," "32K," "Nikon D780," typically all together 🙂 However, most of these words do not have any impact on the generated images. Try using prompts with and without these words, and you'll likely see no difference in the results.

Interestingly, some of these words can even hinder the desired outcome. Including "photorealistic" may lead to the generation of paintings instead of photos, as photos themselves are already realistic. The latest versions of Midjourney generally generate photos by default, but to ensure a photo-like image, simply start your prompt with "a photo of..." – this will suffice.

Furthermore, these additional words overload the prompt and draw attention away from the essential keywords that truly matter.

7. Consider using ChatGPT for inspiration

I understand that some may view this advice as contradictory to the #1 tip of writing short and specific prompts. However, it is not contradictory in practice. The output you receive from ChatGPT largely depends on the nature of your request. For instance, if you simply ask for an image description of a tiger walking down a street, ChatGPT may generate lengthy and largely irrelevant text paragraphs.

So, what is the smarter way to use ChatGPT? In my experience, it serves as an excellent tool for idea brainstorming and language enhancement, particularly for non-native speakers like myself. I usually find it beneficial to request 5-10 concise one-sentence image descriptions, where each description portrays a specific subject in a particular environment, like in the example below.

As you can see, instead of a mundane prompt like "beautiful tiger walking down a city street," ChatGPT generates more captivating and imaginative descriptions such as "amidst the hustle and bustle of city life," "magnificent tiger," and "strolls with poise and serenity." I often combine ideas from various suggestions while ensuring they align with my intended style, removing overly poetic elements when necessary.

However, with further training, ChatGPT can become even more specific and adhere to exact prompt formulas, including starting with "/imagine prompt:" and incorporating the required parameters at the end of the prompt. This will give you the prompts that are ready to be directly used in Midjourney.

8. Experiment with styles, colors, camera angles, lighting

Midjourney offers endless possibilities for creating images in diverse styles, exploring different color schemes, experimenting with camera angles, and applying various lighting effects. Take advantage of this opportunity to enhance your creations and unlock new artistic possibilities that could be very challenging and time-consuming without the assistance of AI.

Good luck and happy prompting!

🗞 News and Top Reads

• Japan’s government reaffirmed that it will not enforce copyrights on data used in AI training. The policy allows AI to use any data “regardless of whether it is for non-profit or commercial purposes, whether it is an act other than reproduction, or whether it is content obtained from illegal sites or otherwise.”

• Microsoft Teams is getting new and free AI tools.

  • Teams now supports AI-generated images in Communities.

  • Moreover, within Microsoft Teams, users have access to a powerful tool called Designer, which functions similarly to Canva. It allows you to generate visually appealing designs for presentations, posters, digital postcards, and more. Designer accepts both text prompts and uploaded images, harnessing the capabilities of DALL-E 2, OpenAI's advanced text-to-image AI model.

• EU urged Google and Facebook to label AI-generated content.

  • In response to the issue of fake news and disinformation originating from Russia, social media companies have faced demands to swiftly implement the practice of labeling content and images generated by artificial intelligence.

📌 AI Art Tutorial: Photoshop has AI

You may be aware that Adobe Photoshop has introduced Adobe Firefly AI as an integrated feature in its latest Beta version. In this insightful overview video, Matt Wolfe delves into its capabilities and showcases the innovative ways others have utilized it. Prepare to be inspired by the remarkable examples of creative expression. Take a look for yourself!

🎨 Featured Artist: Troy Guénard

Troy Guénard is an artist, exploring the true magic of AI and fusing it with his art. Check out his Instagram profile @troy_guenard to immerse yourself in a world of visually striking and elegantly minimalist images crafted with the assistance of AI.

If you want to recommend an artist to be featured in this newsletter, feel free to respond to this email with the links to the artist’s website or Instagram profile. Self-promotion is also allowed 😃 

🖼 AI-Assisted Artwork of the Week

🤓 How to Get Started with AI Art?

1. DALL-E: Creating Images from Text – introduction to text-to-image generation.

2. The DALL-E 2 Prompt Book – a guidebook by OpenAI that explains how to effectively right prompts to generate images across different domains (e.g., photography, illustration, art history, 3D artwork).

3. All-In-One Guide For Midjourney: The Art Of Prompts – a comprehensive guide on creating effective Midjourney prompts by Karol Andruszków.

4. Best Midjourney Prompts – a guide that covers the basics of Midjourney prompts (e.g., which keywords to use to create abstract art, surreal art, minimalism, etc) as well as some more advanced options (e.g., keywords related to camera lenses and filters, imitating certain artists and photographers without using their names). Finally, they provide a list of 600+ creative text prompts for image generation.

5. Stable Diffusion Prompt Book – a prompt book prepared by OpenArt. The book discusses ideal prompt format, using modifiers to change the style, format, or perspective of the image, applying ”magic words” to improve image quality, adding negative prompts, and adjusting Stable Diffusion parameters.

Share Kiki and Mozart

If you like this newsletter and know somebody who might also like it, feel free to share this newsletter. Let’s have more people learn about AI art!

If you have been forwarded this email and you like it, please subscribe below. And welcome to the world of AI art!