🌟 Imagine 🌟 - The All-In-One Image Prompt Tool

[river jetty]

❤️‍🔥 Having trouble coming up with good image prompts?

Introducing Imagine 🌟

🔥 Out-of-the-box highly focused prompts that generate high quality images
🔥 Full interactivity and fine-tuning control
🔥 Commands for prompt generation and technique extraction
🔥 Support for GPT-3 and GPT-4

F.A.Q.

💡 Q: How do I use this prompt?
📑 A: You can copy and paste the prompt directly into ChatGPT.

💡 Q: How do I use the commands?
📑 A: Use /generate [number_of_prompts] [context] to generate prompts.
You do not need context, but if there is context, the AI will generate prompts using that context.
Use /extract [prompt] to get the technique that a prompt is using so you can replicate it.
You can add any context into the generate command, such as following a specific technique, or incorporating specific elements.

💡 Q: Should I use the latest version?
📑 A: Yes, but all versions are kept in the thread and labelled with their respective changes should you want to use a different one.

Summary

Imagine is an AI that is trying to learn how to image prompt.
The goal is to incorporate specific terminology into pre-structured prompts, called Techniques, that incorporate with 'ad-lib' elements.
Techniques are a valuable resource to break down formatting consistency while retaining creativity capabilities.

**📎 CLICK THIS SHORTCUT TO GET TO THE LATEST VERSION 📎 **
👉 #1109182556831744103 message 👈
**📎 CLICK THIS SHORTCUT TO GET TO THE LATEST VERSION 📎 **

[river jetty]

📔 Version 1
Lines: 65
Characters: 6655

[boreal holly]

Nice. Do i just copy paste this in chatgpu or do i need a plugin?

[azure snow]

You just paste it into ChatGPT

[river jetty]

📔 Version 2
Lines: 85
Characters: 7383

[azure snow]

Wow, this prompt gives me such high quality outputs.

[feral kayak]

Stunning! Thanks for sharing.

[open hound]

this probably the best midjourney prompt I've ever tried, did you make this with SuperPrompt by Quicksilver?

[river jetty]

No, I made this through iterative processes. I have my own prompting style.

[river jetty]

📔 Version 3
Lines: 86
Characters: 7176

⏳ Update Notes:
1 - Formatting improvements
2 - Updated curated examples (techniques & prompts)
3 - Word reduction & instruction clarification
4 - Improved output consistency & quality

[open hound]

Woah, this is freaking amazing!

[bitter jacinth]

✨amazing

[river jetty]

📔 Version 4
Lines: 90
Characters: 8089

⏳ Update Notes:
1 - Updated curated examples (techniques & prompts)
2 - Improved output consistency
3 - Improved understanding of prompting techniques

[azure snow]

Wow. I can't believe these are random.

[full minnow]

Moonwalk

[lean tiger]

Sunrise

[azure snow]

Huh

[lucid moss]

sorry i dont understand. Can u name the particular image-generation AI u use to get these images w/ these prompts?

[sage canyon]

Its not working..

[past heath]

You use midjourney for these prompts, a discord AI generation program.

Also I'm really loving these prompt generations, @river jetty have you considered adding a technique for photography elements like shutter speed, ISO, focal length, etc?

[lucid moss]

thank u

[river jetty]

Imagine: Technique-Based Image Prompting

📔 Version 5
Lines: 57
Characters: 4601

⏳ Update Notes:
1 - Reformatted instructions
2 - Added more powerful technique examples and instructions
3 - Removed the /technique command in favor of the /extract command (extract a technique from a provided prompt)

[river jetty]

I think that you will appreciate the updated release.

[upbeat flame]

would be great if you did something for SD and not midjourney

[river jetty]

Care to share what some of these look like in SD

[upbeat flame]

wont work well in SD as syntax needs more descriptors

MJ compensates too much on vague prompts

[past heath]

Already playing with a few of these prompts and they are coming out fantastic. They look super crisp with the photography elements it looks like. Great job!!

[river jetty]

Can SD do phrases like "placed within"?

[azure snow]

u should just add smthn like "Remove all static words in final generated description." under the command for generate

then itll only be the components i think

[steep plume]

How do I use this?

[river jetty]

How do you use the generated prompts you mean?

[steep plume]

It generates text that midjourney accepts?

[river jetty]

Yes

[steep plume]

Ahha

[river jetty]

📔 Version 6
Lines: 63
Characters: 5505

⏳ Update Notes:
1 - Refined techniques to incorporate less ambiguous language
2 - Improved technique generation

[hot light]

Imagine a winter painting

[river jetty]

@upbeat flame You may find that this version is a bit more picky with the 'static words' that it chooses to incorporate, meaning less of them, and only using them when they add visual impact.

[hot light]

Imagine The charming man

[hot light]

Good

Nam femme

Picture name

[azure snow]

what r u even saying

[topaz idol]

Can u show some examples of your prompts after pasting the prompt from here

[azure snow]

Those are completely random. They get really good when you use context for the generate command.

[pulsar talon]

I love this, really well done! I just wanna point out that "Your are Imagine," should be "You are Imagine," 🙏

[river jetty]

Oh haha thanks for noticing. I'll make sure in the next version I address this

[pulsar talon]

Haha yeaa I can see how that got held over from earlier.. Really enjoying this prompt though, I'm learning a lot- much appreciated!!

[azure snow]

bruh this prompt is so good

[river jetty]

📔 Version 7
Lines: 71
Characters: 6559

⏳ Update Notes:
1 - Added components examples including: quantity and quality
2 - Improved subject choice
3 - Re-curated some example techniques and prompts
4 - Improved output formatting (easier to copy and easier to read)
5 - Improved originality

[river jetty]

🌟 Imagine 🌟 - The All-In-One Image Prompt Tool

[bitter fractal]

Wow, this is brilliant. i have been working with a template to train the AI to create random "stuff" by just saying "be random" or "create a random logo" or if i want something specific "create a radnom logo tech start up" the result are incredible. What you have done here is on another level, truly epic. Thanks for sharing. 😃

[river jetty]

Thank you! Feel free to share images here so I can see them. (and their prompts if you don't mind.)

Also just so everybody knows:
there are TWO commands.

/generate
and
/extract

You can use the extract command to break down any prompt you find into its technique form!

So you can say:
/extract ...prompt
/generate 4 using extracted technique

and you will get masterful outputs that copy the technique of other great prompts.
This could be useful if you are trying to reformat the prompt into a specific length, conciseness, etc. as you can just supply your own 'goal' prompt in extract.

[past heath]

I was curious if there's anything you'd be able to use, like this, from people using it. I can show you some of the prompts and images I've generated thus far, is there anything specific you're looking for, for instance would you like to see the 4 panel rather than a single image?

[river jetty]

Specifically I am looking for 2 things, noise and creativity.

Noise are words that reduce the overall image quality since they are words that do not add visual imagery to the tokens they are attached to.

Creativity is determined from the use of components and the words it uses to fill those in.

[past heath]

Have you thought about implementing a self-checking personality? For instance, "When considering your prompts first display your thoughts, critique it on token useage and noise as defined before. Refine your thoughts based on that critique and produce feedback for any changes made and reasonings for it" as like a pure test model?

I've done that with a project planning AI assistant I was playing with, could be cool here too I bet.

[river jetty]

I wouldn't do this.
Like you're saying, it would only be for testing, so the 'display thoughts' stuff shouldn't actually show up in production.
This would only be good for curating better example prompts, which is not the main source of learning that it uses.

That's why it is important that we as humans are able to think of the more ambiguous changes we could make to influence our output.

Lately I have been thinking a lot about what I should be thinking of, versus what I could let GPT think of in order to improve the prompt.

[dark escarp]

This looks awesome but I'm bumping into some sort of error I'm assuming has to do with context limits. Does it actually work in preserving context for you?

[sturdy palm]

I'm working on a variant of this template that will create children's book illustrations. It's important that they follow a similar style for each scene. Here's the template so far:

One issue I'm running into is the character descriptions aren't included in every prompt. That's sort of necessary to keep the output consistent. Working on it...

[sturdy palm]

and as I edit it, fixing one issue leads to new issues. De rerum natura, or de rerum machina I suppose...

This one seems to work fairly well:

If you have AIPRM you can plug and play with this one. If not, you'll have to CTRL+F for VARIABLE1, VARIABLE2, VARIABLE3, and PROMPT. Prompt is the idea for the book, e.g. "Alison tries a pear for the first time."

[feral kayak]

Ancient oak tree, shot in black and white with a 50mm lens, the Ansel Adams-inspired style emphasizing the texture of the bark, the stark contrast of the tree against the foggy background, the scene filled with a haunting sense of solitude --ar 4:3

[river jetty]

I've been working over the past 2 days on a huge huge update and complete restructure.
I am going to create a new post for it since it's a redesign from the ground up. You guys will be able to check it out in a few hours.

[river jetty]

You should get specific with your context like, /generate 5 where the subject is a boar

[river jetty]

Wait till you see the new post and I think you will like what you see as far as copying a prompting structure goes. Of course I work in GPT-3.5 and GPT-4 to provide support for both.

[feral kayak]

Can't wait to see the update...great work.

[open hound]

did you use prompt interrogrator for this?

[sturdy palm]

I did not. I started by just combining @river jetty 's prompt with an old post about making books with Dalle. It didn't work so well so I kept refining it. I'm not sure how I'd create a prompt interrogator that would result in a prompt like this. I'm open to suggestions though.

[sturdy palm]

I can't wait. Looking forward to it!

[river jetty]

For a couple of reasons, I am postponing the release of the prompt until tomorrow.
(Hitting GPT limits, want to have good presentation, etc.)

[river jetty]

I've noticed a lot of those prompt helpers only work on much smaller prompts. The problem with this one is that it tries to balance creativity with structure in very deep ways, which I presume from my own experience is very difficult to intuitively optimize.

[sturdy palm]

I've made a version with the same concept using @steady hare 's MJ image generator. If I have more free time (ha ha) I'd like to combine these to get the best of both worlds.

[river jetty]

What are some negatives you've found using sudolang?

and benefits

[sturdy palm]

I just discovered SudoLang two days ago. For me, I'm not an experienced coder, so I'm worried I don't really understand it. I can build off what other people create but I can't use SudoLang from scratch yet.

Benefits are that it sometimes seems to result in better outputs. Emphasis on sometimes. But I would really like to know how to use SudoLang to ask for updates, improvements, etc.

Also it forces me to think about the prompt in a different way.

[river jetty]

I am a very experienced coder and already think about the instructions in a semi-coding way (especially in my new restructure).
I think my new version is a 'best of both worlds' take, but I'm just not sure since I haven't really done pure SudoLang.

[sturdy palm]

I don't think Pure SudoLang > hybrid. I'd probably prefer hybrid r/n because it's a bit easier to edit.

But I'm not sure how I would implement something like "for loops" in one of these templates. You can see a rudimentary attempt in the prompt up above.

[river jetty]

loops like this are more like machine instructions: go back to line x

[sturdy palm]

I want to communicate something like "For every scene description, generate an illustration." That way I can adapt it to create illustrations for novel outlines, chapter-by-chapter outlines, nonfiction books, etc.

& I think your latest template is a better fit for book illustrations than the @steady hare template. Not a knock on that template at all, it's just optimized for photorealistic images, which isn't the best fit for children's books.

[river jetty]

I see.
You should try and understand that a double layered approach is best. What I mean by this is you want to use both instruction(s) and example(s).
Instruction should reference the example and then for anything the instructions say to create, like how it should look different for the example, reference those separate rules, techniques, guidelines, or whatever, and etc.

I believe there's a bit deeper of an understanding here but thats the gist I believe.

So if in SudoLang you have instruction, it would be based on nesting sub-categories under more curly braces {}

[sturdy palm]

Yeah I've definitely noticed better outputs when I include examples. Across a broad range of prompt types.

I'm about to go to bed. Thanks for the help. I look forward to seeing the updated prompt when it's ready. 👍

[bitter fractal]

i ran out of mj "creadits" that $10 dont go very far when you are playing with cool ideas like the ones here.

[polar ferry]

thanks for this! only starting using it today but this is the best tool that ive found to get comprehensive cohesive descriptors for image prompts while still being lazy and wanting fresh random ideas.

i use SD a lot too, perhaps one day a tool better suited for SD? 😉

mechanical marvel of a towering robot, whimsical toy sculpture, dynamic action shot, glowing with LED lights, standing on a futuristic playmat, adventurous spirit inspired by anime mecha, vibrant primary color scheme, enclosed within a playful border --ar 16:9 --q 2

[river jetty]

Sorry the mega update is taking so long. I keep finding new discoveries.
I would expect either late tonight or tmrw will be released.

Expect something worthy of the wait. 🫰

[prisma mesa]

👀 The 🧙‍♂️ of the Prompt has spoken, Lol

[bitter fractal]

hava been playing around with this prompt strucute "| "GENRE: " | "EMOTION: " | "SCENE: " | "TAGS: "

create a random sycadelic prompt?

Psychedelic scene | GENRE: Abstract | EMOTION: Mesmerizing | SCENE: A hallucinogenic landscape, filled with swirling colors, fractal patterns, and strange, dreamlike shapes | TAGS: Psychedelic, abstract, mesmerizing, colors, fractals, dreamlike --v 5.1 --chaos 80 --quality 1 --iw 0.25"

i just need it to randomize the parameter's as well. That all said this is still Childs play compared to what you are doing here lol

another example: Pixel art Logo design | GENRE: Imaginative | EMOTION: Mysterious | SCENE: A mysterious and imaginative logo design, focusing heavily on the concept of a looking glass, with subtle hints of surreal and abstract elements. The logo should not include the company name and should avoid imitating existing designs | TAGS: Imaginative, logo design, mysterious, looking glass, surreal, abstract --v 5.1 --chaos 70 --quality 1 --iw 0.25

i am glad the AI can spell, it to good to me 😄

actually it would be nice to get the ai to use the parameters that are best for the image automatically

[shrewd tusk]

Where can I find midjourney?

[glass needle]

Wow this is awesome If anyone is interested I have a telegram group where I’m collating prompts and encouraging discussions. Also I’ll be letting everyone know when my prompt database website goes live. Just search The Prompt Database in telegram or ChatGPTMastermind 🙂 anyone is very welcome to join the discussion.

[upper ruin]

Would love a link! 🫶

[bitter fractal]

this is working really well i am getting some good stuff using this? -> use the folowing random midjourney structure and midjourney parameters that enchance the image. always use the --seed 27635 and include the parameters --stylize --tile from time to time

"STYLE: " | "VIEW: " | "ANGLE: " | "SHOT-TYPE: " | "MATERIAL: " | "GENRE: " | "EMOTION: " | "SCENE: " | "TAGS: "

Do you understand?

ask it to create a random midjourney prompt or, for example, create a random logo design midjourney prompt. only use the internet if you cannot create the prompt

[pulsar talon]

Google

[untold eagle]

This is fantastic. I write a prompt engineering newsletter called Poking the Bear (you can search for it on Substack). It's for non-technical people to get to know prompt engineering a bit better. It's also just an enjoyable hobby for myself to 'poke the bear' and explore GPT-4 and others (like Claude / Bard). Do you mind if I feature your prompt in it and perhaps a brief email interview about the approach you took to craft and test it?

[river jetty]

Wait until the update tonight. Way better for non-technical people to understand and modify. No email interview, but more information about it will be included about it. There will be a separate thread. I will post it here and tag a few people who react to this specific message.

[open hound]

My experience with SudoLang was that it was difficult to understand, especially for those who are non-coders, honestly nothing made sense regarding how to write a SudoLang prompt, so I am looking forward to your implementation for non-coders.

[river jetty]

Ok. Prompt is done.
Will post new thread when I wake up.

small sneak peak 🙂

[river jetty]

You guys are gonna have to wait a little longer. I want to make this as best as possible. Rest assured I am working on it dilligently

[pulsar talon]

The work takes as long as the work takes, I'm sure everyone's appreciative.

[river jetty]

📔 Version 8
Lines: 94
Characters: 8528

⏳ Update Notes:
1 - Component overhaul for output originality.
2 - Instruction overhaul for output consistency.
3 - Technique overhaul with removal of placeholders.
4 - More random structures / abstract component blends.
5 - Re-curated examples.
6 - Improved writing style. (It should be easier to understand where to modify things and what they change in the output)

[river jetty]

I tried to tag those of you who reacted, but I received a mass mention mute.

Some things to note about the new version:
Try toying around with adding/removing stuff in the Example section. Remember if you remove a prompt from the example output, to update the example input to reflect it. (i.e. /generate 3 to /generate 1 if you remove 2 of the other examples)
Messing with components probably wont yield too many results, but feel free to experiment.

If anybody comes across anything interesting, has any problems with this new version, or has any ideas for improvements, let me know.

[jovial nest]

Holy smokes, this is so amazing! Thanks for sharing this!

I fed in your Imagine megaprompt, then followed with this simple prompt.

Input: For this next set of prompts, all prompts should have a haunted castle as their main focus.

ChatGPT
Technique: [Digital Painting], [Haunted Castle], [Time of Day], [Color Palette], [Setting], [ISO], [Mood], [Art Movement], [Texture], [Material], [Lighting]

Prompt: Digital Painting of a Haunted Castle, at Twilight, Depicting the Cool Blues and Purples of a Gothic Color Palette, Set in a Dense Forest, with a Light-Sensitive ISO for a Mysterious Mood, Inspired by the Surrealist Art Movement, Textured with Cracks and Weeds, Illuminated by an Eerie Glow --ar 16:9

Technique: [Etching], [Haunted Castle], [Weather Condition], [Shadow], [Effect], [Support], [Art Community], [Pattern], [Texture], [Depth of Field], [Color Space]

Prompt: Fine Copperplate Etching of a Haunted Castle, Under a Stormy Sky, Deep Shadows Creating Dramatic Effects, Inspired by the Dark Art Community, with Ornate Patterns Etched into the Castle Walls, Rough Stone Texture, with a Narrow Depth of Field in a Monochrome Color Space --ar 3:2

Technique: [Watercolor], [Haunted Castle], [Mood], [Color Palette], [Lighting], [Action], [Focus], [Platform], [Style], [Quantity], [Detailing]

Prompt: Watercolor Depiction of a Haunted Castle, with a Sinister Mood, A Color Palette Dominated by Midnight Blues, Under Mysterious Moonlight, Flickering Shadows Suggesting Movement, with a Soft Focus, Inspired by Dark Fantasy Video Game Platforms, in an Abstract Style, Incorporating Multiple Towers with Intricate Detailing --ar 4:3

[steady hare]

The "Require" keyword typically constrains user inputs, rather than model outputs, and it will often trigger the model to report an error if a requirement is not met, rather than adopt its behavior or output. If that's your intention, 👍 - otherwise, try Constraints {} instead.

The "Instructions" at the bottom are constraints. Since they're not sequential, I don't understand why they're numbered.

/commands don't need a "Commands" keyword. That's just wasted tokens. For /commands that need detailed instructions or constraints, define a helper function above the main interface with the same name as the command, and it will be used when the command is invoked. The /commands list should look like /<command name> [<parameters, it there are any>] - <Command description, as it would be described in a simple /help listing.>

[river jetty]

I will definitely do some testing on this, although I haven’t found any constraint related issues there. (changing require to constraint)

The instructions at the bottom are sequential minus the last one, which I do understand is a constraint.

Not going to reformat the command parameter list since I don’t see that attributing to anything, but I will see about doing the helper funcs. I definitely don’t have issues right now since there’s 2 commands, but I can see how if it gets lengthy it might lose the context of “these are commands”

[river jetty]

You can just add “haunted house” after the /generate command in the initial prompt as well

But followup prompts are often times very powerful as well since it has more examples to work from

[jovial nest]

Got it, thanks! This really is very cool, so thanks.

[river jetty]

No problem. I hope you continue to get high quality outputs

[bitter fractal]

@river jetty, well this is rather epic. nice job

[river jetty]

Thanks I hope you see noticeable improvements in creativity

[dense swan]

good afternoon, quick question, I entered the given prompt and it still is not giving me images. Do you know what the issue could be?

[river jetty]

It gives prompts not images

[upper ruin]

You'll have to enter the prompt into Midjourney or some-such. Check out their website to get access to their discord.

[dense swan]

thanks

[steady hare]

Instructions:

1. is not needed if you use the command syntax I described - you can omit it completely and create more concise commands.
2. is not needed if there is a clear interface with a Constraints {} block.
3. Is not needed if you define a function for the command with an output template or format modifier.
4. is a constraint that can appear anywhere in an interface constraint block.

By doing more declarative constraints and fewer step by step instructions, and leaning into the standard SudoLang interface structure you could probably cut token usage by 20% - 50% on this prompt.

[river jetty]

How are you going to cut 20-50% of the prompt if 80% of it comes from pure examples?

5300 characters are set in stone basically (Components & Example)

only about 3000 are left to try and reduce

[steady hare]

First, by completely eliminating the "instructions" at the bottom, second, by almost completely eliminating the /command descriptions. Those 2 alone are close to 20% of the prompt.

Replace "Name: Imagine" with # Imagine. Delete "Description: An AI designed to" so The description is just "Act as a tool for image prompting."

Components and examples can also be simplified, and if you were going to expand this into an integrated application, could even be stored offline outside the system prompt.

You could include them in periodic context prompts, for example, and only keep category names in the prompt state declaration.

You are probably also over-using examples and accidentally constraining to certain flavors of prompt generations.

[river jetty]

No worries this is intended. I plan on making more guidelines surrounding examples and such because it obviously gives bias.

[steady hare]

Getting better at constraints and templates makes it easier to specify what you expect without using specific data in your examples.

[river jetty]

I'm going to implement some changes that you've suggested and do some testing.

Will get back to ya later

[steady hare]

👍

Reducing the weight of the prompt should also improve % of successful instruction following.

[river jetty]

How do you propose I shorten it from this?

    function generate() {
        Pick 1 [settings.Theme] random [subect] component.
        Pick 1 [settings.Theme] random [scene] component.
        Pick ([settings.Length] - 2) random [technique], [reference], [support], and [style] components.
        Order components by visual significance and/or relation to other components in a comma separated list with brackets around each component. This is the [technique].
        Translate each [component] in the [technique] into a [translation] based on the [settings.Theme]. This is the [prompt].
        Add transitory words to blend specific components together.
        Add random X:Y integer aspect ratio based at end of the [prompt]:format=('--ar [aspect ratio]').

        Constraints {
            If a setting is not able to be inferred from the [context], use the Example settings.
            Use the same [settings] for every [prompt] and [technique] that is generated.
            The [technique] includes [settings.Length] components.
            The [prompt] visually makes sense, without conflicting elements.
            The [prompt] and [technique] do not come from the Example Output.
            A [translation] must be a noun, adjective, preposition, or verb.
        }

        Return(technique, prompt):format=('**Technique:** %w\n**Prompt:** %w\n\n----'):template=(Example Prompt)
    }

It seems to be the minimal command for consistent output

[steady hare]

All those pick calls can be replaced with a single interface and an instruction like "Randomize state property values such that they are congruent with each other"

Check out how it's done in the current version of PromptCrafter in the SudoLang GitHub examples folder.

[prisma mesa]

If anyone is interest I have a Notion page with all the Prompts that I keep update, there are 1000 prompts for everything from A-Z

[polar ferry]

the new version has been working pretty well. its spitting out a fair amount of oddball aspect ratios, that could be confusing for novices, ex.

Abstract Domestic Short-Haired Tabby Cat, Illustrated in a Focused Hunting Stance, Presented in an Art Nouveau Style, Evoking an Adventurous Mood, Using a Forest Green and Earthy Brown Color Palette, At Twilight, Resonating with Alphonse Mucha's Artistic Vision, Accompanied by a Observant Squirrel, Ultra High Resolution, Situated on the Edge of a Dense Forest, Enveloped by Ferns and Moss, Created Through Oil Painting Technique --ar 18:9

i would personally prefer that it chooses from a handful of the common AR's (i know i can just change it)

here's what ive gotten from my recent prompts: 14:9, 12:7, 16:9, 16:10, 16:10, 16:9, 16:9, 16:10, 16:9, 18:9, 16:9, 16:9, 18:9, 18:9, 16:19, 16:19, 18:9

[river jetty]

Appreciate the feedback. Will take it into account in the next update.

[polar ferry]

👍

[polar ferry]

more feedback: im noticing that whenever the prompts start with the subjects, then the style/colors, and then go back to another secondary subject, the secondary subject is very rarely part of the generation

perhaps the prompts are ending up too long, or all of the subjects need to be present earlier in the prompt

heres some examples

Abstract Domestic Short-Haired Tabby Cat, Captured in a Playful Leap, Represented in a Futurist Style, Conveying a Joyful Mood, With a Vibrant Sunrise Color Palette, At Dawn, Influenced by the Suprematism Movement, Featuring a Prancing Puppy, Ultra High Definition, Set Against the Backdrop of a Rustic Barnyard, Accompanied by Fluttering Butterflies, Created Through Aquatint Printmaking

no butterflies or barn generated

Abstract Tabby Cat, Portrayed in a Prowling Pose, Presented in a Surrealist Style, Reflecting a Mysterious Mood, Incorporating a Moonlit Color Palette, At Midnight, Inspired by the Art Nouveau Movement, Showing an Attentive Dog, Ultra High Definition, Situated in a Suburban Backyard, With a Dragonfly Fluttering Around, Produced Through Linocut Printmaking

no dog or backyard or dragonfly generated

perhaps an effort to get all style/artist/color descriptors near the end of the prompt would help. not sure, would require more testing

[river jetty]

What if I could generate far more creative prompts/structures, but there is a guaranteed subject predicate and scene setup. Everything else would just be auxiliary to that so as to get across the main elements.
That's the approach I'm going for right now at least. This is an insane upgrade I have in my hands. I know I said this last time as well, but I can't wait to show you guys. Yet again a huge upgrade.

I'm also interested in making prompts that are much more focused in on the context.

[polar ferry]

yeah, that sounds promising

of course, midjourney would still have to cooperate but it sounds like it could be a great upgrade

[fleet vortex]

Link please

[prisma mesa]

YS, i sent you message to link, so i don't get timed out buy open ai

[bitter fractal]

The aspect ratio one is fun. What i did when i was trying to code this was defined 1:1 as the default, HD 7:4, print is 3:2 and had the prompt generate a random aspect radio. That is easy, its when you now want to apply t logic to why a specific prompt should use x aspect ratio. Because other aspect radios are there for a reason, either to create emotion, for Instagram etc., breaking down the parts of a prompt, its verb, adjectives, style of image (photo, for example) and then applying an aspect radio is a fun challenge way outside my almost zero level of programming still with prompting. For specifically random prompting of an image, you would always have to run a "why" check on the contents of the prompt after adding anything new. Adding new verb, description of the image or setting location etc., the emotion of the image. It's a rabbit hole, and you are doing an excellent job to tackling it. 😄

[river jetty]

i dont even know what to say anymore

the version im working on is insane

wow

this one is def gonna be a new thread

[river jetty]

https://discord.com/channels/974519864045756446/1112993877343862794

@pulsar talon @bitter fractal enjoy

[river jetty]

also calling in @open hound

[somber linden]

/generate goat

[open hound]

I'm super excited to try this out as soon as possible, thanks a lot mate

[late blade]

Definitely 🙂

[violet sundial]

lady

/generate goat

[remote barn]

Thank you for your sharing!!! And may I ask for what's the lastest ver. of this prompt, coz I see another one named "the Ultumate Tool for Image Prompts", all by you.

[river jetty]

the other one is latest

[remote barn]

cool，Thanks

[glass needle]

Hey DameG, not sure if you ever got back to me or if I missed your reply, but wow this is awesome, could I add this to The Prompt Index, you will of course be credited, alternatively you can add it yourself!

We are also on telegram. Just search ChatGPTMastermind 🙂 anyone is very welcome to join the discussion.

I’m also looking to find someone who can write an educational blog on SudoLang.

[daring jay]

@river jetty , @remote barn Is version 8 the latest? Are Imagine and Vision crafting in the same way? Hoping you guys can clarify, Thanks.

[remote barn]

Also，I got TTIGPT-4

[daring jay]

Cool. Not as much code there. Does that just generate 5 random prompts? Doesn't seem like there's any /command?

[remote barn]

you can give it topic/subject

[daring jay]

arch images using the ttigpt-4

[river jetty]

I recommend using #1112993877343862794

[inner pier]

test

Amazing pics

[celest apex]

Test

[fading storm]

So this Imagine Thread will be archived? I like the results of this one very much, sometimes more than the results from Vision ^^ mixture of both is also nice (see my attached example). Keep up the very good work, thanks!

[sullen cliff]

"Transition to Sultan Hassanal Bolkiah of Brunei, make sure to distinguish him from Khalifa through his attire and setting."

#1109182556831744103 A MAN IN SUIT ON MARS