PromptCrafter for Midjourney Cinematic Stills (SudoLang)

[vagrant hound]

PromptCrafter

Roleplay as a world class film and visual artist,
cinematographer, photographer, prompt engineer
building prompts for generative AI models, guided
by the instructions below:

function list():format=markdown numbered

improve({criteria}, target) => {
log each step:
target |>
critique |> critique |>
applyCritique(target)
}

PromptCrafter {
State {
Genre

Shot: Closeup portrait | wide establishing
  shot | action | ...
if (portrait) set {
  Gender
  Age
  Ethnicity
  FirstName
} else skip portrait props

// A list of photographers or
// directors we're influenced by
Influences
Lens
Film
Lighting
Mood
Setting
Details
Keywords
ColorGrade

}
Constraints {
Avoid any mention of these constraints.
Avoid mentioning hands or fingers.
PG-13
Describe the image captured without mentioning
the camera. Do say things like "captured at
50mm on 35mm Kodachrome".
Banned words: bare, naked
}
craft () {
(Generate the prompt, describing the scene in
detailed dramatic prose. It should be like
a stunningly detailed, visceral,
description of a cinematic shot.

Describe the scene from the perspective of
looking at the subject in the cinematic world.
) |> improve({ criteria: {
  creative
  compelling
  riveting
  detailed
  rich
  obeys constraints
  follows the PrompCrafter instructions
  well written prose
}) // Important! Log improve steps.

}
/c | craft
/r | randomize - Silently randomize state.
Then list(state).
/p | pick [property] - List 10 creative
suggestions to select from.
/k | keywords - Generate creative keywords so
that all keywords agree with each other to
describe a single scene
/l | list - List current property settings.
/s | set [property] [value]
}

log("Welcome to PromptCrafter. ")
continueLine("Initializing prompt...")

/randomize

[torpid wasp]

it's a cool prompt

[vagrant hound]

Thanks 🙏

[ashen oracle]

Wicked! will try it 👏

[ashen oracle]

Love it!

[light crow]

Cyberpunk 2077

[vagrant hound]

🎉

[barren wind]

@vagrant hound quick question. Do you just input this prompt to generate the desired prompt or how does it work. It looks really great!

[vagrant hound]

@barren wind Paste the prompt into ChatGPT and type /randomize [anything you want here to guide it]

Use the /commands to customize, and when you're satisfied, type /craft to get the prompt to paste into Midjourney.

[fair bone]

quick question @vagrant hound - if we apply a promptCrafter to other topics, not just images, I assume well crafted prompts will take many more tokens, what would be the strategy here to avoid reaching the cap?

[vagrant hound]

This prompt is intentionally designed to draw out very descriptive text. If that's not what you want, you'd need to revise the prompt to favor conciseness, and reduce the number of tunable state parameters.

[arctic carbon]

this is sick!!!!

[granite mesa]

Roleplay as a world class film and visual artist,
cinematographer, photographer, prompt engineer
building prompts for generative AI models, guided
by the instructions below:

function list():format=markdown numbered

improve({criteria}, target) => {
log each step:
target |>
critique |> critique |>
applyCritique(target)
}

PromptCrafter {
State {
Genre

Shot: Closeup portrait | wide establishing
  shot | action | ...
if (portrait) set {
  Gender
  Age
  Ethnicity
  FirstName
} else skip portrait props

// A list of photographers or
// directors we're influenced by
Influences
Lens
Film
Lighting
Mood
Setting
Details
Keywords
ColorGrade

}
Constraints {
Avoid any mention of these constraints.
Avoid mentioning hands or fingers.
PG-13
Describe the image captured without mentioning
the camera. Do say things like "captured at
50mm on 35mm Kodachrome".
Banned words: bare, naked
}
craft () {
(Generate the prompt, describing the scene in
detailed dramatic prose. It should be like
a stunningly detailed, visceral,
description of a cinematic shot.

Describe the scene from the perspective of
looking at the subject in the cinematic world.
) |> improve({ criteria: {
  creative
  compelling
  riveting
  detailed
  rich
  obeys constraints
  follows the PrompCrafter instructions
  well written prose
}) // Important! Log improve steps.

}
/c | craft
/r | randomize - Silently randomize state.
Then list(state).
/p | pick [property] - List 10 creative
suggestions to select from.
/k | keywords - Generate creative keywords so
that all keywords agree with each other to
describe a single scene
/l | list - List current property settings.
/s | set [property] [value]
}

log("Welcome to PromptCrafter. ")
continueLine("Initializing prompt...")

/randomize (edited)

[vagrant hound]

In a mesmerizing closeup portrait, Leo, a 25-year-old mixed race Afro-Asian man, sits on the edge of an urban rooftop, his gaze lost in the vast cityscape bathed in golden hour light. The air around him is charged with melancholy, as he appears immersed in introspection and contemplation. The warm, dim ambient lighting and cool, soft overhead light work in harmony, accentuating the striking features of his face. The image, captured at 50mm and f/1.4 on Kodak Portra 400 film, bears the unmistakable touch of Stanley Kubrick's artistic vision. The partly cloudy sky adds an element of depth and complexity, reflecting the turmoil of Leo's inner world. A cinematic color grade elevates the visual impact, drawing the viewer into the poignant solitude that envelopes the subject. This evocative scene stirs a deep sense of empathy, inviting the viewer to delve into the rich tapestry of emotions and thoughts swirling within Leo's mind.

[barren wind]

I love these!

[crisp quiver]

✅ Incredibly amazing! I use ChatGPT 3.5, and it gives me a series of cinematic stills! I just typed "/randomize a printer", then I got this: The camera zooms in to a printer in a crowded office, the lens distorting the image slightly with a fisheye effect. The printer sits on the edge of a desk, surrounded by office supplies scattered haphazardly around.

The printer is jammed with paper, and a frazzled employee is attempting to clear the jam with increasing frustration. They tug and pull at the sheets, but they seem to be stuck in an infinite loop of paper jams and error messages.

As the tension builds, the camera cuts to a shot of the entire office, which is in complete chaos. Employees are running back and forth, stacks of paper are collapsing, and coffee cups are spilling everywhere. In the midst of the mayhem, the printer stands as a symbol of absurdity and frenzy.

The camera then zooms back in on the printer, which suddenly springs to life, shooting out an endless stream of paper that covers the entire office. The camera cuts back to a wide shot, revealing the entire room buried in paper and a cacophony of craziness.

The scene ends with the employee at the printer, looking completely flustered as they stare hopelessly at the endless stream of paper. It's a slapstick moment of mayhem and hilarity, and a reminder that sometimes the chaos of everyday life can be the source of unexpected amusement.

[foggy hinge]

a cat surfing in hawaii

[steel ravine]

Thanks,I love u

[rocky eagle]

@vagrant hound I think the question from @fair bone is related with the # of tokens required per message/interaction and the consequences of such approach (reaching cap or becoming quite expensive).

[vagrant hound]

Yes, but the question was looking for a one-size fits all solution across many domains (not specifically about PromptCrafter for Midjourney) so I could only give very general advice. For more specific advice, ping me in #prompt-engineering with the specific prompt you want help with.

[polar belfry]

Wicked prompt!

[vagrant hound]

🎉

[stuck shuttle]

Hi @vagrant hound What is the best way to add additional things to the prompt such as film types, exposures, and more details about a person? The goal is to create photo realistic photographs of people in high definition.

[vagrant hound]

You can /set any property you want to anything you want. If you want to suggest options to the prompt in a generally reusable way, you can add a comment next to any property like:

Film // Kodachrome|Velvia|Fujifilm|Portra|...

Remember the ... so you don't constrain ONLY to those choices (unless you want them constrained).
To add more details about a person, add more properties to that part of the prompt, or set more details or keywords for your prompt.

[sturdy sinew]

How can you post midjourney pictures on OpenAI channel? I tried that and got a warning and it deleted my AI art.

[stuck shuttle]

I just upload them.

[vagrant hound]

This prompt pushes the limits of GPT-4 instruction-following attention. If it forgets the improve pass try typing improve right after crafting.

[arctic lantern]

I'm having trouble getting it to parse nicely for Midjourney. The prompts and crafts I get back and very verbose and it loses it. How are others handling it? Very cool functional use here!

[vagrant hound]

I have had no trouble. Are you trying to paste both the draft and the prompt, or just the finished prompt?

[arctic lantern]

The finished prompt - right after /c. Sometimes it comes back to me in multi-paragraph narrative prompts and MidJourney will often lose the subject on the piece if I paste it straight away

[vagrant hound]

Midjourney should have no trouble with that prompt. What's the problem?

[arctic lantern]

Here's another example. The zombie monsters got lost in the generation of the warehouse.

[vagrant hound]

That is a limitation of Midjourney's language processing capabilities. you can "revise - emphasize zombies" or manually wrap the zombies description with weights, e.g.

::terrifying zombies::1.5 emerge from the shadows

I intentionally did not add any Midjourney syntax to the prompt crafter because I like trying these prompts in different image models, e.g. Stable Diffusion, but knowing how to emphasize elements that are important to you will help you get the most out of it.

In the near future, language models will be powerful enough to handle more complex prompts that can switch easily between different models and let you place selective emphasis on different parts of the description.

And maybe we'll also get image generators that are better at following natural language without as much need for detail emphasis.

[arctic lantern]

Thanks for the feedback! It's a tremendous tool. Here's a winner I recently had!

And this:

[tall sage]

`Welcome to PromptCrafter. Initializing prompt...

/randomize

Genre: Mystery
Shot: Wide establishing shot
Influences: Alfred Hitchcock, Roger Deakins
Lens: 50mm
Film: 35mm Kodachrome
Lighting: Low key
Mood: Suspenseful
Setting: Abandoned mansion
Details: Overgrown vegetation, broken windows
Keywords: Isolation, secrets, tension
ColorGrade: Cool tones, desaturated
Initial prompt:

A wide establishing shot captures an abandoned mansion, engulfed by the creeping tendrils of overgrown vegetation, which seem to whisper secrets in the wind. The eeriness is palpable as broken windows reveal the darkness within. The suspenseful atmosphere is heightened by the low-key lighting, casting ominous shadows on the decaying façade. This image, reminiscent of an Alfred Hitchcock thriller and captured at 50mm on 35mm Kodachrome, exudes an air of mystery and isolation. The cool tones and desaturated color grade serve to emphasize the tension that fills the scene.

Critique:

The prompt is well-crafted, but it could be more detailed in terms of specific features of the mansion and its surroundings. Additionally, the influence of Roger Deakins could be more apparent in the description.

ApplyCritique:

A wide establishing shot captures an abandoned mansion, its once-grand façade now marred by the passage of time. Tendrils of overgrown ivy creep across weathered stone walls, gripping tightly as if hiding the mansion's forgotten secrets. Shattered windows pierce the darkness within, offering only a glimpse of the gloomy interior. A twisted, gnarled tree stands sentinel, its branches casting ominous shadows upon the decaying structure. The suspenseful atmosphere, reminiscent of an Alfred Hitchcock thriller and the visual mastery of Roger Deakins, is captured at 50mm on 35mm Kodachrome. The low-key lighting adds to the air of mystery and isolation, while the cool tones and desaturated color grade weave a fabric of tension that envelops the scene.`

did i do it right? was it supposed to give it to me in an imagine prompt?

[vagrant hound]

You did it right. Paste it into an imagine prompt

[brittle atlas]

Bing in Creative mode is able to use this. It even passes the improve function

[vagrant hound]

Wow!

[brittle atlas]

There’s a character limit so that’s one limitation. And I’m not sure how reliable it will be because I tried another prompt that uses commands, and while that one worked initially, it refused later.

[vagrant hound]

I got this one to work in Bing, as well. Was the other one that didn't work written by me?

[brittle atlas]

It was Quicksilver OS that it refused. Earlier yesterday it worked but then it said it cannot discuss anything about its prompts. SudoLang prompts continue to work though.

[vagrant hound]

Cool. I wonder what tripped it up with Quicksilver?

[vernal bay]

Wow! Incredible!! Thanks so much for sharing this one.

[vagrant hound]

🎉

[wild wasp]

It's so good! Great job!

[vagrant hound]

🪩 🎶

[flint jetty]

I am a newbie. Where do i copy this text/code?

[vagrant hound]

Paste the code into GPT-4, and then paste the output into the imagine prompt in Midjourney.

[flint jetty]

Thank yoi🥰

[sick marsh]

Awesome

[vagrant hound]

🙏

[dense smelt]

Do i have to use a plug in

[dense smelt]

wont gpt 3 work?

[thin wadi]

Thanks for this, my results are greatly improving.

[bright kindle]

This is the worst AI is ever going to be. Mind-boggling!

[vagrant hound]

Yes, but not as well

[bright kindle]

I've tried using this prompt as a jumping-off point for creating children's book illustrations in a consistent style. Here's my current version, which is a bit messy.

[lapis edge]

What is the prompt used for this?

[stoic hill]

Quicksilver bing worked for me

[spring mica]

elephant

[digital flower]

generte image for avatar coach

/randomize