Prompt Design Resources for MidJourney AI

by Shane McGeehan

Blog: Recent Entries

Newest Posts

Bangin’ Prompts

Is this a joke? Obviously, yes.
Does it kind of make you think about things? I think so.

I’ve been pondering prompting a lot recently, and how sometimes it doesn’t matter if your prompt is short or way too long, has commas or parentheses, contains nonsense, or is even incredibly well thought out with detailed terminology – it almost doesn’t matter, because you WILL get some results that are interesting. AI is just really good at making an intriguing composition.

Folks who don’t like AI think you can bang your head on the keyboard and make “art” and do they have a point? Yea, kind of. You can (literally) smash a bunch of keys, spell things wrong, not actually know what a word you used actually means, and in the end you might get an image that is incredibly beautiful or surreal… but that’s not what we do, well, most of us.

Don’t get me wrong, I love silly gibberish prompting for fun, but the real work in AI is experimenting with structure, terms, and ideas to get something we envision, or explore things we can’t. I can bang my head on a keyboard, and I did, but control and exploration are what I am really after – and that does take a lot of time and skill. There’s no denying that.

(click the prompt to copy it!)

/imagine prompt: nby hjugt76 vcfd m ,jkn vc v --v 5

(click to view)

Describe, Or Don’t

The developers just added a really fun (and useful) command: Describe! Where MidJourney is mainly a text-to-image AI, the new /describe command now let’s us turn image-to-text, in a prompt-like format. This can be fun for just playing around, but it can also be useful to find new terms for images you like but can’t seem to describe… but keep in mind, it’s not always to be trusted.

As a photographer, I have been quite outspoken against many camera terms (like aperture or ISO settings) because they never seem to do anything logical, so there is no use in typing them – however – I am getting a lot of camera types and settings back from the new /describe feature. So this is a good reminder to take what it says as inspiration, not as some master prompter.

Okay, so I just uploaded an image of R2D2 on the beach and used the /describe command, and this was the first prompt that came out. I really think it was onto something with the repetition of “pop-culture-infused” (sarcasm) so I decided to ad a ton more to make my prompt even more amazing. 🤣

(click the prompt to copy it!)

/imagine prompt: r2 d2 standing with a beach chair under an umbrella, in the style of historical reimagining, pop-culture-infused, pop-culture-infused, pop-culture-infused, pop-culture-infused, pop-culture-infused, pop-culture-infused, pop-culture-infused, pop-culture-infused, pop-culture-infused, pop-culture-infused[…]

(click to view)

The Film

People think AI is simple and you just type stuff – sure, sometimes. I have been working on this prompt for 5 days trying to get what I wanted, and still didn’t quite get 100% there, but I decided to use what worked. This new piece is called “The Film” and is just a short about pain.

I’ve been listening to a lot of Nine Inch Nails lately, which isn’t my norm, but he’s some great tracks. The young gamer in me really loved the Quake soundtrack that Trent Reznor did. Music inspires, and so I had this idea to create an old film movie showing pain, suffering, saddness. The dark stuff.

I used the new V5 with a weird aspect ratio of 1:7 to get film strips that I was able to zoom into. I then upscaled the imagery with Visions Of Chaos, animated the frames in Adobe Premiere, and added some sound mixed from Riffusion AI and my own edits in Adobe Audition.)

(click the prompt to copy it!)

/imagine prompt: IMAGEURL film strip scan of a man crying screaming in the rain --ar 1:7 --stylize 1000 --chaos 0 --v 5

(click to view)

Fonts In MidJourney

Months ago I came up with a method to spell words in MidJourney using image prompt, but it takes a lot of time because you need to design a letter in Photoshop, save it, upload it, and prompt it – for every single letter! I had an idea to automatically batch an alphabet of prompts, and after some testing – it works!

The next version of Prompter will include a new FONTS batching feature. All you have to do is type a style (or pick some from prompter’s many menus), pick a bright or dark background and BAM, prompts for every single letter are ready to copy and paste – each with individual URLs and text automatically added.

I don’t know a release date yet, still have a lot of coding to do, but it’ll be free and a lot of fun! Here are letters in the style of neon lights, sketches, colorful fur, and the man himself – Mr Bob Dylan. 

(click the prompt to copy it!)

/imagine prompt: IMAGEURL the letter F photograph of Bob Dylan holding a sign with a letter on it --v 5

(click to view)

Don’t Use GTP

A little warning on this post. Usually I am a nice and helpful guy, but time to time I need to go on an angry rant, lol… GPT is an amazing AI technology, and incredibly helpful for everything from communication to coding, but it is NOT for MidJourney. When I posted about using GPT with MidJourney, it was an attempt to “train” it on syntax, parameters and such, and I envisioned a speech-to-prompt tool, but by no means am I saying we should use it for the main prompt.

There is this new trend of people using GPT to write MidJourney prompts, and it makes me laugh. I just can’t take the idea of people using a long book-like paragraph from GPT as a prompt. Did the image come out cool? Of course it did. You can bash your head into your keyboard and the images will look interesting. Anyway, I decided to use GPT as well, with some dark humor and sarcasm…

(click the prompt to copy it!)

/imagine prompt: Wibbledy wobbedy floo, a bumbleflap zapperdoodle swooped through the air, leaving trails of shroomfizzles and whizzing sneezles. A grand gooblesnort, with its wobbly chumblywump, danced atop a flibberdijibbet, spinning dinglebops and tingly spangleblams. In the midst of all the blopitty blurps, a whooshing sound swept through the confuddled jibberjabber, as the clitterclatter of zibberzabber filled the air with a wafting scent of zumblequacks and frizzlefrazzle. The skriddlydoo wobbled and wiggled, its magnobberous tendrils tickling the wigglywaggly fribbledyflop. Abruptly, a fandangled zorklewooper erupted, spraying giddybits and spronklewinks in a dazzling display of jumblejazzle. And there, in the midst of the kerfluffledy goo, a shimmering flibberdegibbet whispered, "Shplonk!" --ar 16:9 --q 5 --v 5

(click to view)

Willow Dog Runs

I have a little corgi puppy that I adore, but she is a bit crazy. We like taking her for hikes in the woods, but she seems to think the point of hiking is to run at full speed, like it’s a race, gotta win! It makes me laugh, and so I wanted to make a flicker animation of her running day and night.

I first used an image prompt of her face with a description, re-rolled a few times, and found one I felt really looked like her. I then used that MidJourney image in another prompt to get the woods scene I wanted. After finding the image that had a good look to it – corgi, motion blur, paws, trees – I re-rolled over and over, upscaled the images, then created an animation (blending some frames together to ease the flicker feel). The rest was some simple color grading that fades from bright sunlight, to orange glow, some purples, then night… To all of you it’s just a cute running dog, but I laugh because this is 100% my Willow Dog.

(click the prompt to copy it!)

/imagine prompt: https://s.mj.run/sHP5JxWpo9k cinematic fisheye action shot blurry legs young happy furry tricolor cogri running in the forest bokeh --ar 16:9 --v 4 --style 4c --v 4

(click to view)

Tutorial: Batcher, For Batching

So I just made a new tool in Prompter called… Batcher… and It does exactly what you’d think, you know, it batches, duh. If a prompt builder is designed to create one prompt, why not have one to create TONS of prompts to pick from?! Typing sucks.

Check a box next to medium, or color, or camera. Batch a huge list, or pick and choose by hand selecting terms you like. It’s easy, and since it’s built off of Prompter, you have a big database of terms to test out, and can even add weights, stylize, chaos, or aspect ratios.

With Batcher you can turn something like a “decrepit mansion” into a plethora of styles: brutalism, low poly, psychedelic, horror, stone age, world war, fractal, kaleidoscope, ice and rain – see what works, and upscale the best!

(click the prompt to copy it!)

/imagine prompt: diorama decrepit mansion silhouette:: ice::1.2 --seed 420 --stylize 700 --chaos 10 --ar 1:1 --v 4 --style 4c

(click to view)

ChatGPT Personality Portraits

I had an idea the other day. If ChatGPT is an AI that is like a person who always does as they are told – what if I told them to be multiple people instead. I began by creating some personalities, such as CHEERFUL MAN who is always upbeat and happy, or DEPRESSED MAN who has a cynical outlook on everything. I then added more, like STONED MAN who is high all the time, or POETRY MAN who is a romantic who loves to rhyme (among others too). It started out fun, getting all kinds of answers and opinions rather than just one from GPT, but then I went deeper.

I started to have them talk to eachother about topics. They began to develop life stories, some with creative upbringings or even abusive parents. I incerted myself into the chats and STONED MAN got upset because he works in a grocery store while I get to work from home… It’s incredibly fascinating, actually.

Anyway, I then “taught” GPT what MidJourney is and how the syntax works, parameters and such. I thought, what do these people look like? So I asked them all to write prompts describing themselves and their setting. Their prompts were awful. Like, bro, that’s not how you write a prompt (but we all know GPT ain’t made for that thing). Still though, it was incredibly fun, and actually interesting to not only see their personalities develop but then an image of them. Below is the prompt from POETRY MAN as an example.

(click the prompt to copy it!)

/imagine prompt: A serene and picturesque park, with a tranquil pond and lush green trees. You see a man with a contemplative expression, wearing a tweed jacket and corduroy pants. His hair is gray and his face is lined with wisdom. He is holding a leather-bound notebook and a fountain pen in his hand. The beauty of his surroundings inspires him to create poetry that captures the essence of life. --v 4 --ar 9:16

(click to view)

Strange Planet in 360 Virtual Reality

I am a bit of a virtual reality geek. There is something so surreal or trippy about being able to stand in a small room yet feel like you are flying a jet around New York City at sunset, or sculpting creatures with digital clay. I foresee a future where we can blend AI and VR in real-time, and literally create landscapes and characters around us. I can’t even imagine how fun that could be once the tech catches up with our imaginations.

I had the idea to create a 360 panorama long ago, but never had much luck with V3. I began using V4 to create wide shots and heard someone use the term “equirectangular projection” which I had never heard of. It was perfect… I came up with a workflow to create multiple images in MidJourney, blend them together in Photoshop, then copy part of the beginning back to the end, and blend that so that I get a seamless 360 image. I then used a tool to embed 360 camera metadata so that I could post it online. Lastly, I had to fire up the old VR, launch Virtual Desktop, and stand inside my chaotic and dismal alien planet (along with some weird mixed audio I made in Adobe Audition).

(click the prompt to copy it!)

/imagine prompt: equirectangular projection of a alien UFO in a landscape and purple volcanoes lava and green stormy skies lightning seamless insanely detailed ultra-wide angle lens --stylize 500 --ar 2:1 --v 4 --style 4c

(click to view)

Tutorial: Prompter and Randomness

I’ve been wanting to make some videos, and I figure, with a new version of Prompter out, what a better time than now. Finally you all know what my voice sounds like. All natural. And no, it’s not one of those fake AI voiceovers, lol.

No one likes really long videos, so I wanted to keep this short. I talk about why I use Prompter, some great features you might find helpful, and we also play with the newest feature – RANDOM. Thanks to ChatGPT, an idea I had long ago is now a reality. Let’s get random!

(click the prompt to copy it!)

/imagine prompt: logo of a rocky mountain surrounded by fog fluorescent tube dark-orange --ar 16:9 --v 4 --style 4c

(click to view)

Blog Search

Blog Categories

Blog Archives

USA

This is a piece I began a little while ago and finally felt the need to finish, given the recent...

CUBE³

I began this project a while ago (back in version 3) but set it aside while working on new ideas....

Cursed

You wanna know a secret? I tried to keep quiet about this for weeks, but since someone else let...

Melting AI

I am a photo artist. Much of my work begins with a camera, but it isn't your typical photography....

Christmas Cards

I decided to make my own Christmas cards with MidJourney AI this year. To be honest, I rarely ever...

Pin It on Pinterest

Share This