Advanced Midjourney Prompt Craft Workflow: Welcome to Wonderland
Generative art tools like Midjourney do an amazing job of creating eye-catching imagery with very little manual direction. However, to consistently create stunning images in Midjourney you’ll need to develop a process that relies less on Midjourney “magic” and more on repeatable steps that help you improve your prompt engineering.
Sifting through mountains of example prompts or scanning pages of image thumbnails can be a time consuming approach. Not to mention, working with lots of unfamiliar creative jargon really doesn’t give you a ton of control over your creative output.
Rather than simply showcasing effective prompt elements, the goal of this guide is to provide a flexible prompt crafting workflow that’ll help you develop detailed prompts quickly and consistently. This workflow will include a few 3rd party tools — all free to use — that will help increase the speed and quality of our prompting output.
Now that we have the introduction is out of the way: let’s move past the looking glass and deeper into wonderland territory.
Getting started: setting up shop
Here’s an outline of the workflow we’ll be working with:
concept -> prompt generation -> prompt selection -> prompt refinement -> finished image output
This workflow will take us from rough concept to detailed prompt to eye catching image output. To follow along with this workflow you’ll need access to the following:
- Midjourney: https://www.midjourney.com/home/
- ChatGPT: https://openai.com/
- and the browser extension Harpa AI: https://harpa.ai/
Harpa AI is a free extension you can use to run custom ChatGPT prompts right in your browser. It’s one my go to AI powered productivity tools. Harpa also happens to include a helpful Midjourney prompt generator that does a great job of creating initial draft prompts to work with.
We’ll be sure to include image and prompt examples along with a few tips and tricks to help customize this workflow for your own Midjourney creative projects.
Drafting prompts: the magic starts here
We’ll start with our concept and aim for something simple. We want our concept to be little more than a subject and some detailing to describe the look and feel of what we want.
For the examples in this guide I decided to re-imagine a few of my favorite characters from Alice in Wonderland–the Queen of Hearts and the Cheshire Cat. Here’s the starting concept I’ll be working from:
Queen of hearts from Alice in Wonderland, voodoo queen, dark, moody, cinematic
Selecting the Midjourney prompt generator in Harpa AI we’ll have two options: quick and custom. We’ll select custom and Harpa will ask us for the following inputs: [concept], [style], [details], and [aspect ratio (ar)].
We have our concept ready to go. The style we’ll input is “photorealistic”. Photorealism is a great style to use with this workflow prompts can often include lots of specific detailing — camera models, lens types and shot angles. We’ll leave details blank and choose an aspect ration of 1:1.
By leaving the details section blank we can have Harpa/Chatgpt help provide lots of useful detailing for us. Our square 1-to-1 aspect ratio is a great starting point we can easily tweak later on in the workflow.
Prompt generation and selection
Harpa will spin up 5 detailed variations of our prompt for us. We’ll take these prompts over to Midjourney and see what our initial outputs look like. For each prompt we’ll include the following extra instructions/parameters at the very end:
--style raw
--repeat 3The “style” command tells Midjourney to pay more attention to our prompt detailing and to add less of it’s own Midjourney “magic” to the mix. The “repeat” command tells Midjourney to spin up 3 batches of 4 for our prompt. Running multiple generations of our prompt helps speed up our selection process and gives us a better sense of the direction the prompt output is going in.
We’re looking for a prompt that gives us 4–5 images out of a set of 12 that have the right look and feel. Once we have a prompt that feels like a good fit we’re ready to move on to the next stage of our workflow.
Here’s the initial example prompt from Harpa along with the extra Midjourney parameters I choose to refine in our next step of the workflow:
The Queen of Hearts, reimagined as a voodoo queen, sits on a throne made of twisted vines and bones in the middle of a swamp. Her expression is stern and regal as she holds a scepter with a heart-shaped gem. The environment is dark and atmospheric, with the dense foliage of the swamp creating a natural canopy overhead, and the flickering light of torches casting long shadows. The style is photorealistic, capturing the texture of the throne, the queen’s attire, and the surrounding foliage. The scene will be captured with a Nikon D850 camera and a Nikon AF-S NIKKOR 50mm f/1.8G lens, using a wide aperture to create a shallow depth of field, focusing on the queen while softly blurring the background --style raw --repeat 3
Here are a few examples from our initial example prompt:
These initial image generations helped me better visualize my concept. I like the idea of a darker version of the Queen of Hearts — fueled by the earthy magic of powerful Loa Spirits.
Prompt refining techniques
The next tool we use in this workflow to refine our prompt is Midjourney’s “shorten” command. The “shorten” command helps us better understand how our prompt is being evaluated by returning our original prompt with some styling to let us know which parts have a high, moderate, and low level of impact on our image output.
We’ll also receive a few modified versions of our prompt paired down to just the most high value elements we can use for image generation right away.
Here’s our prompt along with some image examples after using the “shorten” command on our original prompt:
The Queen of Hearts, reimagined as a voodoo queen, sits on a throne made of twisted vines and bones in the middle of a swamp. Her expression is stern and regal as she holds a scepter with a heart-shaped gem. The environment is dark and atmospheric, with the dense foliage of the swamp creating a natural canopy overhead, and the flickering light of torches casting long shadows. The style is photorealistic, capturing the texture of the throne, the queen’s attire, and the surrounding foliage. The scene will be captured with a Nikon D850 camera and a Nikon AF-S NIKKOR 50mm f/1.8G lens, using a wide aperture to create a shallow depth of field, focusing on the queen while softly blurring the background
Our high value elements are in bold, moderate in plain italic, and the low impact elements as underlined text/links.
Using the “shorten” command to rework and remove elements from our original prompt, the image examples above display a better meld of character motifs and vampy voodoo chic.
The temptation here in this workflow is to simply toss away all the elements that don’t impact our output — just working with everything else. For some projects this might be the best choice to make.
However, this approach also means you’ll miss a golden opportunity to take the wheel and really drive your prompt crafting workflow. Using commands to change Midjourney’s “mind” about which parts of a prompt should be more or less important is a great way to steer the direction of your creative output.
Checkout this guide on how to add “weights” to your prompts in Midjourney.
Midjourney commands like “shorten” and “describe” are amazing resources to better understand how Midjourney “thinks” about the prompts and images we provide.
After using “shorten” to evaluate our original prompt you’ll see I made quite a few adjustments. You’ll also notice that the main elements of our original concept are still alive and well in both the finalized prompt and the image output.
But don’t take my word for it — see for you self.
[finished prompt]
Photorealistic image of a happy Queen of Hearts:5 radiant beauty:5 adoring smile:6 decorated by red heart shapes:4 alice in wonderland:3 luxurious voodoo queen:2 sits on a throne made of twisted vines and bones in the middle of a swamp:2 expression, delighted, inviting, warmth, environment is dark and atmospheric, with the dense foliage of the swamp
These image renders above do a great job of capturing our concept of a vibrant voodoo queen, flaunting her favor and dominion for all to see.
Not to be out done. Here are a few refined output examples using this workflow to give the Cheshire Cat some seriously sinister form. The concept at work behind these image gens below re-imagines our fabled feline as an avatar of elemental chaos — at the peek of predatory power.
Workflow crafting tips
- Photorealistic and CGI generative art styles often benefit from detailing that describes the camera, lens, view angle or rendering engine used to create them
- Prompt generators are a great time savers for creating highly detailed prompts to start experimenting with quickly
- Experiment with more or less Midjourney “magic” by toggling the “raw” style parameter to see what works best for your creative project
- Leverage built in Midjourney features like “Shorten” and “Describe” to better understand how Midjourney “thinks” about prompts and images you input
- Make a cheat sheet of effective prompt elements for certain styles to speed up your creative process
- Use the “shorten” command in Midjourney to evaluate your prompt elements and help decide what part to refine
- Use the “describe” command in Midjourney to work backward from image generations to craft more effective prompt elements
- Generative art tools allow you to experiment with an impressive variety of creative elements quickly
- See my guide on how to break up prompts into “blocks” that can help with the refinement process
- Check out my guide on prompt weighting techniques
The goal of this guide is to share a workflow for creating eye-catching art in Midjourney that you can tweak and make your own. Along the way, I hope you found something useful — something fun, or just kind of cool — to help take your own Midjourney creative process to the next level.
Stay tuned for my upcoming prompt crafting workflow guide for Leonardo AI…coming soon!
Happy prompt crafting out there, cheers!
