Author: Lars Nielsen
Translator: Chen Chao
Proofreader: Zhao Ruxuan
This article is approximately 1500 words long and is suggested to be read in 5 minutes.
This article lists 10 tips on how to write image generation prompts for Midjourney.
A detailed “cheat sheet” along with some keywords for better image outputs using prompts.
Image generated by Midjourney, text prompt: A beautiful, dreamy city in Unreal Engine
For those who haven’t heard of Midjourney, here’s a brief introduction: Midjourney is a text-to-image generation app, similar to OpenAI’s DALLE-2 and Stable Diffusion’s DreamStudio, using a vast amount of online images (approximately 650 million) to generate stunning images based on provided text clues. It is currently in a testing phase, but more features are added every week… and the future is yet to be imagined! (Currently, their AI engine is at version 3.0).
Those who have used Midjourney know that the more adjectives you use in the prompts, the more vibrant and unique the output becomes. Users typically use raw and direct prompts: “A cowboy in a tuxedo on the moon,” which will generate the following pattern:
Image generated by Midjourney — Prompt: A cowboy in a tuxedo on the moon
However, there are many more options to create predictable and consistent image outputs than just plain raw text prompts!
Now let’s dive into the advanced options for text prompts!
Here are some ways you can generate variations of images, along with some advanced settings: (For detailed settings, please follow the table below)
1. Provide keywords — ‘style’
6. Pass images as prompts via URL
7. Weight images as prompts
In summary, your style is influenced by the keywords you choose — and you can specify what kind of style you want to achieve — broad style categories are shown in the image below —
Style categories in Midjourney — Image from the author
Providing a set of prompt keywords related to “style” can lead to a series of different outputs based on your preferred style. Here are some keywords and the art forms/designs/artists/styles you may want to choose.
-
Use design/genre as style
Image generated by Midjourney, prompt from the author
-
Use artist names as style
You can also use an artist’s name as your output style. Here is an example of the same prompt generating different outputs:
/Imagine painting a horse running at sunset in the style of Andy Warhol
Image generated by Midjourney, using the artist as style, prompt from the author
Image generated by Midjourney, prompt experiment from the author (Prompt text: “Dreamy Castle”)
You can add settings –s <some number> (indicating style)
The following images are from the same prompt — with high and low stylization options added.
/imagine firefighters --s 6000
Image generated by Midjourney, prompt: firefighters
3. Chaos — Increase Abstraction
Increase or decrease the level of abstraction of the theme within a range of 0-100
/imagine Eiffel tower --chaos 60
Image generated by Midjourney, prompt: Eiffel Tower
To specify the output resolution, you can choose general keywords such as 8K, 4K, photorealistic, ultra photorealistic, ultra detailed, intricate details, etc.
Alternatively, you can standardize settings for predictable outputs using hd and quality / –q as two such settings
/imagine red rose flower --hd/imagine red rose flower --quality 5
You can specify the width:height ratio of the output image. This is called the aspect ratio. The default output is square (1:1 aspect ratio). However, if you want a cinematic perspective — or if you just want to make a wallpaper for your computer, you can adjust the aspect ratio.
/imagine jasmine in the wild flower --ar 4:3
If you want to specify custom image dimensions, use the following examples:
/imagine jasmine in the wild flower --w 600 --h 300
You cannot customize the aspect ratio — but you can specify some standard ratios and some non-standard ratios! (Here are some examples)
Image generated by Midjourney, text prompt: “Wild Jasmine”, text from the author
If you want to get outputs similar to the style of a certain image (when you want consistent output), pass the image URL in.
/imagine http://www.imgur.com/Im3424.jpg box full of chocolates
The generated image will take cues from the seed image (the URL you provided) and the text prompt.
-
You can upload multiple images as prompts
-
You can specify the weight of images (see the next item)
7. Weight Images as Prompts
If you want your output to look more like your prompt image (see the previous item), give that image a higher weight:
(keyword : --iw <number>)/imagine http://www.imgur.com/ks34f24.jpg chocolates --iw:4
/imagine wild animals tiger::2 zebra::4 lions::1.5
9. Filter Words from Images
Use the —no keyword to remove unwanted subjects from the image:
/imagine KFC fried chicken --no sauce
10. Finally, you can try some fun keywords
* Sony Alpha α7, ISO1900, Leica M, = to specify any lens type or camera type* photorealistic, ultra photoreal, ultra detailed, intricate details, = to specify some possible details and a realistic look and less artistic* unreal = to specify an unreal engine feel* volumetric light, cinematic lighting = to specify some lighting conditions
We discussed how you can control the output images to a certain extent in Midjourney experiments. But it’s not finished yet. There is still a lot of imagination and experimentation space. Try different prompts and enjoy your creations, good luck!
Creativity in Text-to-Image Generation JONAS OPPENLAENDER, University of Vaasa, Finlandhttps://arxiv.org/pdf/2206.02904.pdf
To try some text prompts, here’s an article(https://pythoslabs.medium.com/lessons-learnt-from-using-midjourney-v4-and-some-awesome-images-generated-text-prompts-included-2153dc02db4a), which contains some extraordinary images generated by Midjourney – some sample images.
An advanced guide to writing prompts for Midjourney (text-to-image)
https://medium.com/mlearning-ai/an-advanced-guide-to-writing-prompts-for-midjourney-text-to-image-aa12a1e33b6
Chen Chao, Master of Applied Psychology from Peking University, data analysis enthusiast. Previously studied computer science as an undergraduate, and later pursued psychology. During the learning process, I increasingly realized the wide application of data analysis, hoping to output meaningful work through what I’ve learned, and very happy to join the Datapi family, staying humble and eager.
Translation Team Recruitment Information
Job Description: Requires a meticulous heart to translate selected foreign articles into fluent Chinese. If you are a data science/statistics/computer major studying abroad, or working overseas in related fields, or confident in your foreign language skills, you are welcome to join the translation team.
You will receive: Regular translation training to improve volunteers’ translation skills, enhance awareness of cutting-edge data science, and overseas friends can stay connected with domestic technological development. The THU Datapi industry-academia-research background provides volunteers with good development opportunities.
Other Benefits: You will have peers in the translation team from renowned companies’ data scientists, Peking University, Tsinghua University, and other prestigious schools.
Click on the end of the article “Read the original text” to join the Datapi team~
Reprint Notice
If you need to reprint, please prominently indicate the author and source at the beginning of the article (Reprinted from: Datapi ID: DatapiTHU), and place a prominent QR code for Datapi at the end of the article. For articles with original identification, please send [Article Name – Pending Authorization Public Account Name and ID] to the contact email to apply for whitelist authorization and edit as required.
After publication, please provide the link back to the contact email (see below). Unauthorized reprints and adaptations will be legally pursued.
Click“Read the Original Text” to embrace the organization
