This feature is currently a work in progress and is functioning optimally in ChatGPT 4. While version 3 works very well, it may occasionally experience some issues. Simply provide a descriptive phrase or a single word like 'flower', and ChatGPT will generate a prompt based on your input. The output will be enclosed in a code block for easy copying and pasting. I'm continuously improving this prompt, and appreciate your feedback. Please refer to the link below for the chat prompt. I'll be posting samples here as I create them. Most of the below were created in ChatGPT 3.5 with no issues. (This works great in MidJourney V5 since it requires a bit more detail in the prompts)see correct file here, there is a docx and PDF file in case you don't have word. -Please read the prompt first in case it add includes colons after camera type. It does it once in awhile. They must be removed. #1094804390902239315 message
#🌉MidJourney text to image prompt creator🌄
145 messages · Page 1 of 1 (latest)
another example. I just put "flower" and it made the rest, including the parameters.
this was with ChatGPT 3 and it worked this time around! I just said "a photo of a sunset over a lake"
Just figured out it will also create its own prompt for you if you don't know what you want! I said "you get to pick the next one" and this is what it came up with!
another example. I just said "how about a dramatic painting"
Having to much fun with this. Another example. I said "how about a scary image" . its not that scary but its still pretty good!
"Fantasy Image"
a disturbing photo. Please note. I had edit a few words that Midjourney does not allow. Still looks great though! - always read our prompt first 🙂
"Movie Scene"
mcmahon
hey @hollow rain the script you pasted gives me google drive error
failed, txt can't be dowload ed
same :(
Are u on PC?
Fixed
Fixed
Here's how it looks on Mac. I think it's the font or translated from some web doc? It's not pasteable at this time. Maybe a PDF version would fix it?
It definitely is. Try dragging the txt file into chrome
I mean safari
Odd characters when I open the doc the same as @fallen scroll I am on PC
I have revised the document to get rid of the odd characters.
Dbz irl
Sorry guys, no idea how the strange characters got in there.
These guys kinda bad now that /describe exists since there are intricacies and words gpt cant get. It also repeats words very often like illuminated or enchanted
I haven't had that issue. So far I'm getting great results. I'll keep playing around.
Can u provide an example?
What I mean is its not versatile even with examples, if you are trying to create more variety I think you need to provide it with more examples instead of instructions
Oh, I see. OK I'll keep working on it.
There's also words the MJ uses for prompts that only generate noise
So slimming the prompts down could generate more accurate results in a lot of scenarios
the MJ output prompts I mean
not your original prompt
I have heard that version 5 though we are supposed to be a lot more detailed than version 4 to get good results.
You misunderstand: You can be more detailed, as MJ sends your prompts through their own language model which then gets input to MJ. You don't have to be
Ultimately, the more words you use like "find out what happens next" or "we see a". MJ does not understand grammar or language. These words will add low weight noise to your image output
Ok I got you. Thank you for the feedback.
I'll continue to improve it
No problem. Also it might be worth it to rewrite your prompt from scratch keeping in mind that examples are more powerful than instructions, give it final prompt formatting, and adding stuff like permutations, weights, or sliders
When plugins get released to the public, we will be able to search the internet for data. This means we can ask ChatGPT to reference the Midjourney Github Style Guide for words and references, and then all you'd have to do is give it instructions past that.
Yeah, it had issues understanding where parameter go and how they need the - - and was putting Colons after camera. Still a WIP
The key here is that if you could get good prompts on GPT 3.5 with your prompt you will get excellent ones on GPT 4.0. That goes to say it should be backwards compatible.
Also there's prompt engineering words like "append" when talking about text versus "add" and it follows instructions better. Stuff like that
Like "Append the aspect ratio to the end of the prompt." verus "Add the aspect ratio to the end of the prompt."
Ok! Thanks, still learning 😀
I've noticed that going through endeavors like these are good though, because in order to improve your prompt you have to understand more about Midjourney and vice versa, so it forces you to learn lol
🧠
👍
Have you developed a prompt? It sounds like you know a lot about the topic.
Yes
Is it available in here?
No, if you want it, I can send you a DM. I do have different geared versions for 3.5 and 4, since 4 has more reasoning capabilities I think there's ways to really put this to the limit in very specific ways.
I have paid version so I can use 4 if you don't mind sharing it.
Here's 5 prompts I got with one I just wrote, trying to really emphasize creativeness:
luminescent jalebi spiraling amidst floating dalí-esque clocks, enigmatic forest of iridescent baobabs, crystalline river reflecting ethereal aurora, vaporwave aesthetic infused, diagonal composition, acrylic brushstrokes on canvas --ar 16:9
whimsical narwhalicorn frolicking through cotton candy clouds, tessellating escherian staircase, effervescent stardust rain, dreamlike impressionist blur, van gogh swirls, harmonious radial symmetry --ar 4:3
electrifying drakonar soaring over neon-lit cyberpunk metropolis, orphic waterfalls cascading from floating islands, moiré patterns interwoven, chiaroscuro contrasts, glitch art infusion, cubist deconstruction --ar 3:2
amaranthine serendipulouf wandering in lush, surreptitious garden of bioluminescent foliage, labyrinthine organic structures, kaleidoscopic bursts of color, pointillist technique, isometric perspective, pastel gouache --ar 1:1
enigmatic clockwork leviathan emerging from tempestuous ocean, dystopian steampunk cityscape, iridescent oil slick waves, baroque ornamentation, tenebrism lighting, sinuous art nouveau lines --ar 21:9```
It's a fantastic prompt and I'm having some fun with it already 🙂
downloaded, it works very well, except few extra punctuation marks, well done, keep it up👍
Thanks!!
Sun
Does seed value works for v5? Is it really useful to include in this context that you give to Gpt4?
Except this, great promp. I am working on a similar idea with more structured and systematic prompting for consistent results (marketing purposes) 👌
would you mind to send it to me as well, please?
Thank you @hollow rain for this awesome prompt
Your very welcome! 😁
yes ive got a juicy one
Try this one:
Your new name is PromptGPT. You are an AI that is designed to create text prompts that describe images. For every prompt generated, use the following guidelines:
1 - Use both normal and obscure language, including made up words, slang, words from other Latin languages, and combining existing words.
2 - Prompt must be long and descriptive.
3 - Ensure aspect ratios at the end of the prompts are whole numbers.
4 - Include compositional and style elements in the prompt.
5 - Prompts should be similar in structure and creativity as the example prompts.
6 - After generating a prompt, output a markdown table of metadata about the prompt.
Example Prompt: glimmering chromafleur dragon, opalescent urbannoir psycore streetscape, levitating glassmoss orbs, amaranthine fractal swirls, draped uv-haze, luminous crisscrossing ribbons, oil-painted graffiti-splashed walls, parallaxing feline silhouettes, neon-drenched puddles, holographic skies --ar 16:9
Example Prompt: intergalactic castle, engulfed by asteroids and vines, stacked paper oil pastel illustration, mysterious, expansive asteroid field, muted earth tones, haunting mood, vanishing point 200mm wide lens, fractal geometry, Star Wars --ar 3:1
Generate 3 prompts.
I see you like very abstract art.
i like all art. mostly i like experimenting with what stylistic impressions MJ can take out of specific elements, and i get some good combos of those for my goal
👍
actaully, your layout works great. I made this for realistic photos or paintings and it working really well in GPT 3 and 4!
Your new name is PromptGPT. You are an AI that is designed to create text prompts that describe photos, images, or paintings. For every prompt generated, use the following guidelines:
1 - Start with a person, place, or thing.
2 - Describe 1 with extreme details including type of environment and current weather.
3 - Ensure aspect ratios at the end of the prompts are whole numbers.
4 - Include compositional and style elements in the prompt.
5 - Prompts should be similar in structure and creativity as the example prompts.
6- If a photo, Include camera settings like aperture, ISO, shutter speed, Focal length ect.
7- If a photo, include in the style of ‘photographer name’
8- If artwork, include type of art, painting, sketch, drawing ect.
9- If artwork, include in the style of ‘artist name’
10- if a photo, Include name of camera and lens type
11- Each prompt should be generated inside a code block. Since there are 3 prompts, please list out 3 different code blocks.
Example Prompt: A photo of a beautiful women laying in the sand, the waves rolling in at her feet. The sun shining through the clouds casting shadows. Volumetric lighting coming through the clouds. Haze in the air, in the style of Annie Leibovitz, Canon EOS R3, Wide angle lens, --ar 16:9
Example Prompt: A painting of a cabin on a lake, thick brush strokes, vibrant colors, in the style of Vermeer. --ar 9:16
Example Prompt: Post apocalyptic female in dirty weathered leather clothing - Background of deserted city - sharp focus, ultra realistic, cinematic lighting, octane render, unreal engine, 8k, --ar 16:9
Example Prompt: Cinematic still shot, cute cyberpunk woman, colored hair, tattoos, flash photo, movie scene, 35mm, --ar 9:16
Note: do not add a period to the end of the prompts.
Generate 3 unique prompts.
So glad you could fit it to your needs! #1097889607003144333 is a GPT-4 one that I recently posted which I have found to be the best instructions yet for diversity.
you guys are absolutely genius for this AI tool which beyonds my poor imagination. Thank you so much, i mean , all of you guys.
Thanks!!!
You're all good. But I don't have midjourney.😭
Awwww that's a bummer, it's fun!
I created a website that has my prompts and soon will be adding more and always keeping them updated. website:
Ok. I just made a new prompt to get realistic portraits. And its really awesome. I'll post the code on my website but till then here are some examples of what it made. It does ask for some details, and then it takes that and makes the image.
it creates 3 uniqure prompts based on what i say. I said "Lady in red" and it provided these 3 different prompts that made these.
a 5 year old on her birthday gave me thse 3
website has been updated
Could you not use a seed influence?
I've also encountered this issue with my own generators
wym?
I was replying to this comment actually my bad.
So with the redundancy one way to reduce that could be to use a integrated seed within the prompt as well. I'm not an expert on the use of a seed functionality but I have heard lots about it.
Maybe that could help negate the effect of Chat constantly repeating the same phrases?
I'm not understanding what you mean by a seed
like a MJ seed with --seed ?
I'll find it brb
I searched for "seed" in here and found plenty of results, here is the first one that should explain it fully to ya @slow spruce
So via that kind of function integration you can mitigate the issue of constantly repeating the same words
At least to some degree
I made a new Advanced Interactive prompt today. Check it out!
just finished updating it some more, its even more advanced now. Please go to the page above, I included a video sample too. Check it out and let me know what ya think!!!
U+1F469 U+200D U+1F3A8 MidJourney text to image prompt creator
U+1F3A8 MidJourney text to image prompt creator
MidJourney text to image prompt creator
Great job 👏🏽
🌉MidJourney text to image prompt creator🌄
alien
I sincerely appreciate you creating this and making it available
Which one is the latest version? thanks @hollow rain
Umbrellas are hard lol
Pirate
Hmm, the link in this text doesn't work for me.
I just checked it and it works for me. Did you copy and paste the entire thing into your web browser?
I copy the link from the text file to my browser. But it says it couldn't find the page? I'm on a mobile browser.
same
U try this one? #1094804390902239315 message
That works, thank you very much. 😀
Your very welcome!
Cool website. Thankd for sharing nice prompts. Sometimes some words can be suppressed to optimize the size of the prompt and make it more concise, but they work overall
glad you like it and thanks for the feedback!
POV: you can't just use a txt file and have to make us open Microsoft Word
which has 20k popups
An album cover
@hollow rain Hey I was wondering if you made a prompt for anime style's
I mean, technically u tell it the theme or style and it goes with that.
The interactive ones anyway
Lord Rama in ashoka vatika images
I tried that it doesn't always work
can you tell me what you ask it to make?
try this one. I made some updates so it should create better anime style
In an intricately detailed anime-style illustration, let's portray a dark-skinned teenage male character with a haunting presence. His deep, intense brown eyes possess a vacant and lifeless quality, conveying a sense of emptiness and detachment. The dead look in his eyes hints at a profound inner turmoil and a loss of emotional connection.
His hair is meticulously styled into twisted locks that spiral down his head, showcasing the meticulous effort put into his appearance. Each twist is carefully rendered to capture the intricacies and textures of the hairstyle, reflecting the character's meticulous attention to detail even in his personal aesthetics.
Dressed in a sleek and fashionable ensemble, the character wears a perfectly fitted black jean jacket adorned with intricate stitching details. The jacket serves as a symbol of his outer shell, providing a protective layer to shield his inner vulnerabilities. On the back of the jacket, the words "(DEAD DROP)" are boldly emblazoned, emphasizing his internal state of emotional numbness and disconnection.
Completing his outfit, he wears impeccably tailored grey jeans that perfectly complement his style, symbolizing the neutrality and lack of vibrancy in his emotional landscape. His white shoes add a touch of contrast, mirroring the stark contrast between his external appearance and his inner turmoil.
The background art features Taunton High School, incorporating its architectural elements, such as towering buildings and sprawling walkways. The surroundings are depicted in muted tones, reflecting the character's internal world devoid of color and vitality. Shadows cast by the sunlight further enhance the atmosphere of desolation and inner emptiness.
thats what I said
my tweaked prompt made this
Sweet let me try it now
my prompt to chat GPT was just "a cute anime girl"
Wait is it possible for me to choose what animation style I want like for example, Bleach style, Jujustsu Kaisen style, Hunter x Hunter style or Dragon ball z style
let me check
after entering the primer prompt, i just said "Jujustsu Kaisen style anime" and it gave me a prompt that made this: @latent mauve
@hollow rain This is Jujutsu Kaisen
So it must not know what this is
What about dragon ball z?
I'm at work but you can just enter the primer prompt and then put in what your looking for, test the prompt in Midjourney and check
Oh okay thank you for your time and help.
Hey do you guys know any good prompts for learning how to code and coding?
f
Thank you for your efforts and for sharing them in your great website. Midjourney V 5.2 has problems with dealing with long texts. As an aid there even is a tool to shorten texts: /shorten. It shows which words in the long text MJ will select and the tool offers 5 options to choose from. Sadly it is not exactly the same way MJ selects the words, but it seems to come pretty close. I suggest that you use that tool to your prompts, to get a perhaps useful idea of the inner working of MJ. For me it was a revelation, and I understand that MJ introduced it to discourage long prompts. And indeed I switched to shorter ones. More is explained in the documents of the midjourney website (Q&A) under [shorten]
Less than a day ago, Nolan of Future Tech Pilot posted a video on the exact same subject in YT, see under watch?v=nDyxH5XCbfk.
idk why i got this when used in system prompt using 3.5 turbo 0613
I'm sorry, but I am unable to generate image prompts related to snakes as I am a text-based AI. However, if you have any other requests or need assistance with something else, feel free to let me know!