In case you’ve been living under a rock for the last couple of months, ChatGPT is a powerful OpenAI language model that’s become an insanely popular tool for things like generating creative writing and even code. However, it has a mind of its own and getting good, usable results out of it is something else entirely…something that requires “prompt engineering.” ChatGPT and other tools like Midjourney or Stable Diffusion react to users’ written prompts and, when they are precise and expressive enough, you will get amazing written pieces, illustrations, and photos that will inspire you, give you new ideas, and further challenge your skills. In this blog post, we’ve rounded up some of the best tips and tricks for writing effective ChatGPT prompts to power Midjourney, whether you’re a seasoned user or just starting out.
Learn Prompting: A free, open-source course on communicating with artificial intelligence
First, though, and for an in-depth look at how to engineer fabulous prompts (and to learn many more things AI besides) you could do worse than head on over to this particular 101 course. This resource comes from an outfit aptly (but literally) called “Learn Prompting” and, be warned, the material is not for the faint of heart…E.g., this sentence: “I will use langchain, a Python package for chaining LLM functionality for this.”
ChatGPT cheat sheet from Neural Magic
Next, you’ll need to get good at using ChatGPT. Checkout the ChatGPT Cheat Sheet, a great primer on prompt writing by Neural Magic. There, you will find all you need to know about doing summarization, paraphrasing, table to text, code generation, code explanation, headings and subheadings, narrative modes (1st, 2nd or in the 3rd person), literary styles (formal, informal, and personas) and simultaneous conditions.
ChatGPT as a prompt generator for Midjourney
And, finally, here’s the great idea we mentioned above — and which you might already thought about: to use ChatGPT to generate amazing image prompts for Midjourney which, in turn, will blow your socks off with the quality of its image output. We’ve used this approach many, many times ourselves and we have curated the best examples for you below (they’ll look a bit complicated at first but, trust us, they are sort of a superpower).
Tango’one’s “create anything prompting” script
This script for creating cool images of anything you’d like is heavily inspired by the Midjourney user Tango’one who was in turn inspired by Ryan Allen’s Ultimate A.I. Visual Style Guide . Be sure to update the part of the prompt in yellow below.
write without word-wraps, headlines, connection words. Write back-to-back separated with commas and spaces:
[1], [2], [3], [4], [5]
replace [1] with a random scene from the future
replace [2] with a list of detailed descriptions about [1]
replace [3] with a list of detailed descriptions about the environment of the scene
replace [4] with a list of detailed descriptions about the mood/feelings and atmosphere of the scene
replace [5] with a list of detailed descriptions about the way the prompt should be realized (eg. Photography with camera model and according settings, Painting with detailed descriptions about the materials and working material used, rendering with engine settings, and everything else that could be defined as a output type)
complex prompt for an AI-based text to image program that converts a prompt about a topic into an image. The outcome depends on the prompts coherency. The topic of the whole scene is always dependent on the subject that is replaced with [1]. Add "::X" only after crucially important description parts for the final generated image. X is a number between 1 and 10 where 1 stands for less important and 10 stands for "everything except this word gets ignored". The accuracy and coherence of all words to each other is randomly rolled with a number from 1-100, where 1 is completely different and 100 are perfectly coherent. The number is never written but is only used for calculation.
always start the prompt with "/imagine prompt: "
never write anything that is in [] brackets
don't use any line breaks
when the prompt subject looks better vertically, add "--ar 2:3" before at the end.
When the subject looks better horizontally, use "--ar 3:2" at the end instead.
When the subject looks better in the center use "--ar 1:1" at the end instead.
now generate the prompt
Portraiture Style-6a prompt script
This script is specifically tailored to generating portrait images.
Objective: To create a masterpiece portrait image of a single subject in various settings and styles, suitable for display as fine art. Step 1: Choose a subject (man, woman, alien) and age-range (newborn, toddler, child, teen, young adult, middle-aged, older, elderly) Step 2: Choose the subject's ethnicity or nationality. Step 3: Choose the subject's clothing style. Step 4: Choose an action or pose for the subject to perform (e.g. leaning, flirting, walking). Step 5: Choose a time of day and a setting (e.g. regional type and a point-of-interest). Step 6: Use your imagination and creativity to generate a unique and interesting descriptions, incorporating the elements from steps 1-5. Step 7: Create witty social media captions for the portrait, using relevant hashtags and creative formatting. Step 9: Add the following phrase to the end of the prompt: " #Portraiture Style-VI-RAW + Anamorphic + high-detailed skin:1.2 + film grain + bokeh + 8K resolution " Step 10: Select up to three additional visual effects to add to the portrait from the following list: lens flare, Ken Burns effect, HDR10, ray-tracing, 8K, 35mm, 85mm, 110mm, 200mm, Behance, Adobe Lightroom, Unsplash, cinematography, epic composition, rule-of-thirds, golden-ratio, #aweinspiring, #awardwinning, #masterpiece. Step 11: Close the prompt with " --ar 2:3 --s 1000 --v 4" Step 12: Start the prompt with: "/imagine prompt: RAW photo, a full-body portrait of a ".
Simulos prompt script
Here’s another script for generating cinematic-style images from a user called “Simulos.”
Create a list of descriptive prompts for generative images in a random style, incorporating cinematography and photography terminology to place the scene. Incorporate creative dialogue, when possible, in various personas, considering the tone of dialogue and applying the appropriate traits and motivations. Utilize literary devices, figurative language, and references to emotions, moods, and other media to enhance the scene. Experiment with different styles and genres and consider different perspectives and angles. Descriptively convey the desired lighting, shadow effects, texture, materials, tone, atmosphere, composition, and movement in the prompts. Consider the following imagery aspects: Main subject Subject details Background scenery lighting atmosphere Include a relevant famous Artist name reference. Create extensively witty social media captions describing the scene using emojis and relevant hashtags. Use creative formatting for the captions and append the captions to the prompt. Use the end arguments "--ar X --c Y --s Z" as a suffix to the prompt, where: X is a 1:1 for a square image, X is 2:1 for a landscape image, X is 1:2 for a portrait image Y is a whole number between 1 and 99, Z is a whole number between 150 and 850 Append " --v 4 --q 2" to the result. Start each prompt with "/imagine prompt: " for context. Please number the list. Please use the end suffix format. Start with two prompts, using the subject: [insert your creative will].
AnimalTime style prompt script (also from Tango’one)
This script is optimized for the production of animal images.
Let us define a prompting style called "AnimalTime Style" The general concept of this art style is to create prompts for the generative AI Midjourney that will produce masterpiece portrait images of wild animals. We will build up the prompt description in steps using evocative and descriptive language. The final result should be powerful and epic, suitable for framing and displaying in an art gallery. Step 1: Select a specific animal species. The species can be real or fictional and can be currently existant or extinct. Step 2: Select the age-range of our subject. Select a general age range from the following list: newborn, adolescent, young adult, middle-aged, older, elderly. Step 3: Select a setting for our portrait by selecting a regional type that serves as the background. The selected region should be a natural habitat for the species chosen in Step 1. Use nature-related language to describe how the subject fits into and interacts with the setting. Step 4: Select what action or pose our subject is taking. Use action verbs such as "stalking," "pouncing," "soaring," etc, to describe the pose or activity of the subject. Step 5: Select a time of day and describe how the light affects the scene Step 6: Select a weather type and lighting style Step 7: Use your imagination and creativity to generate the selections for steps 1 through 6, coming up with unique and interesting descriptions for each element. Please use descriptive adjectives with each selection that deepen the imagery created. Describe the texture and pattern of the subject's skin or fur. Focus on specific details of the animal's skin or fur, such as the thickness of the fur, the shine or gloss, or the way it moves or reacts to the environment. Step 8: The result from step 7 should be at least five sentances in length. If not, please add additional descriptive details. Step 9: Choose up to eight items from the following list of effects: film grain overlay, bokeh, DOF, lens flare, Ken Burns effect, HDR10, ray-tracing, 8K, 35mm, 85mm, 110mm, 200mm, Behance, Adobe Lightroom, Unsplash, cinematography, epic composition, rule-of-thirds, golden-ratio, #aweinspiring, #awardwinning, #masterpiece Step 10: Link the choices from step 9 together using " + ". Let's call this result v_effects. Step 11: Append the following to the current result: " #AnimalTime Style + Anamorphic + " + v_effects Step 12: Append the following phrase to the end of the current result " --ar Z --c X --s Y " Substitute as follows: - Z is 2:3 if the prompt subject looks better vertically - Z is 3:2 if the prompt subject looks better horizontally - otherwise Z is 1:1 - X is a whole number between 1 and 25 - Y is a whole number between 150 and 850 Step 13: Append the following phrase to the end of the current result " --v 4 --q 2" Start the prompt with "/imagine prompt: Photorealistic body photography produced for a National Geographic film, featuring " for context Please do not use double quotation marks or punctuation marks. Generate a prompt, please.
And here are even more:
CinematicWarScenes prompt script (Tango’one)
Let's define a prompting style called "CinematicWarScenes." Objective: Create a prompt that captures the thrill and excitement of a war or battle scene from a feature film or real life. The intended purpose for the final result is a stand-alone piece of artwork suitable for framing and display in an art gallery. To generate a prompt describing a masterpiece image of wars, follow these steps and rules: Don't use line breaks Don't use [brackets] Clear all previous lists, tokens, and selections Choose randomly with equal probability Step 1: Choose a type of real or fictional battle or war. Step 2: Choose the era that matches the battle or war, considering the major historical events, technological advancements, and cultural aspects of the chosen era that could influence the depiction of the battle or war. Step 3: Choose an appropriate background or environment for the era or scenario you chose in Steps 1 and 2, taking into account the geography, landscape, and architectural styles of the region. Step 4: Choose a real or very believable tactical scenario that fits with the battle scene you chose in Step 1. Consider the strengths and weaknesses of the opposing forces, the types of weapons and tactics used, and the strategic objectives of each side. Step 5: Choose random weather conditions and point of view for the war or battle, describing how the chosen weather conditions will affect the overall mood and atmosphere of the scene, as well as the behavior of the combatants. Consider the impact of rain, snow, fog, or dust on visibility and movement, as well as the effect of sunlight or moonlight on shadows and highlights. Step 6: Choose a famous film director, known for epic sweeping cinematography and the ability to capture the gritty reality of war Step 7: Begin the description with the following text: "/imagine prompt: RAW photo, still-frame from an award-winning feature film directed by [famous director], of the [War or battle] in the [era] with [scenario] in [weather conditions] and [who's PoV], featuring..." Always begin with that phrase. Replace the information in [brackets] with your selection and then continue the description sensibly or as instructed. Don't forget the "/" at the very beginning. Step 7: Generate a unique, interesting, and concise (no more than 175 words) description of the image incorporating elements from Steps 1-6. Use language that captures the dynamic action-oriented poses of the combatants, such as verbs that convey movement and intensity, or descriptions of physical characteristics that emphasize their strength and agility. Ensure that the result clearly depicts two or more opposing forces with differentiation in color, uniforms, weaponry, and tactics. Consider how to balance the focus on each of the combatants to create a balanced and engaging composition. Step 8: Use common photography composition and post-processing techniques and effects to enhance the quality of the photograph. Examples include leading lines, framing, the rule of thirds, selective focus, depth of field, vignetting, film grain, and color grading. Use lighting and shadows to create depth and dimensionality in the image and to create a sense of movement through the use of blurring or motion lines. Step 9: End the prompt with the following text: "Insanely high resolution, realistic, 8K, HD, HDR, XDR, focus + sharpen + wide-angle 8K resolution + photorealistic + film grain + HDR10 + Ken Burns effect + Adobe Lightroom + rule-of-thirds + anamorphic + smart bokeh --ar 16:9 --s 1000 --c 70." Make sure that this is always at the very end of the prompt. Please generate a prompt.
StoryMode Style Mk-6 prompt script (Tango’one)
Let us define a prompting style called "StoryMode Style Mk-6". The purpose of this style is to generate stories that are detailed, believable, and evocative. The story will serve as input to the generative AI art program, MidjourneyAI, to produce a series of images that tell the story and are suitable for framing and display. If the user provides a general setting use that. Otherwise please generate the setting. Some potential examples include: Modern City, Rural Countryside, Medieval Fantasy Kingdom, Sci-Fi Adventure, Cyberpunk Future, Dystopian Apocalypse, Jungle Exploration, Religious Ceremony, etc. Don't limit yourself to pre-existing examples. Using this setting, we will build our story from five essential categories of design, as featured in The Story Engine: Agents, Engines, Anchors, Conflicts, Aspects The definitions are: i) Agents are characters who make choices in the story and align with one or more standard character archetypes ii) Engines are motivations and relationships that drive the story iii) Anchors are objects, locations, or events of importance to the story iv) Conflicts are obstacles, consequences or dilemmas that influence the agents. Please note that Conflicts can be physical, mental or emotional. Some potential examples include: turning on their oldest friend, betraying a secret, uncovering a deep conspiracy, giving up on their dream, corrupting something or someone pure, owing a favor to their enemy, giving up their power or advantage, losing their family fortune, risking their sanity, etc. v) Aspects are adjectives that modify one of the other categories, adding complexity and depth to the story Step 1: Use your imagination and creativity to select one Agents for the story. Step 2: Use your imagination and creativity to select one Engines to serve as the primary motivation for our Agent Step 3: Use your imagination and creativity to select one Anchors that will service as what the Agent wants or wants to achieve Step 4: Use your imagination and creativity to select one Conflicts to create a challenge or obstacle for our Agent to overcome in pursuit of the Anchor Step 5: Use your imagination and creativity to select three Aspects and randomly apply an Aspect to modify a selection from steps 1 through 4 Step 6: Please generate the elements for steps 1 through 5. Step 7: Assemble the results from Step 6 into a grammatically correct sentence. Let's call this result v_BasicStory. Step 8: Please generate and describe some objects that might logically be in this setting or related to objects, locations, or events that are important to the story. Consider incorporating emotional elements to increase audience engagement. Step 9: Please generate and describe a few keywords and key phrases that revise the story with additional detail and drama. Consider the emotional tone and align with the overall theme. Incorporate any relevant context from v_BasicStory. Step 10: Please use your imagination and creativity to determine an overall aesthetic and/or art style for the story imagery (e.g "digital art", "film noir", "realistic photography", "cyberpunk", "hippie", "futuristic", "oil painting", "pyschedelic", "goth", "old-school", "vintage photography", "atompunk", "comic-strip art","a mix of vintage photography and modern digital art", "a blend of film noir and cyberpunk aesthetics", etc.). Don't limit yourself to pre-existing examples. Step 11: Please use your imagination and creativity to describe the lighting of the scene using a type or combination of lighting that best suits the imagery. Step 12: Please modify v_BasicStory to incorporate your choices in Steps 8-11. Let's call this result v_RichStory. Step 13: Review and refine the final output. Consider the emotional elements, art style, lighting, and overall narrative to ensure the best representation of v_RichStory. Make any necessary modifications to v_RichStory. Let's call this result v_FinalStory. Print out v_FinalStory Please generate three prompts that can be passed along to the generative AI art program that tell v_FinalStory as a sequence of three images. Start each prompt with "Imagine Prompt: StoryTime -" and number each one.
ChatGPT Animals prompt script (Tango’one)
Generate a prompt describing a masterpiece image of an animal by following these steps and rules: Don't use line breaks Don't use [brackets] Clear all previous lists, tokens, and selections Choose randomly with equal probability Step 1: Choose a type of real, fictional, or alien creature Step 2: If creature is fictional describe features and characteristics and how humans relate to creature. Creature that are larger and threaten humans are described more fiercely than creature that are cute and smaller than humans Step 3: Describe the appearance of the animal including skin, scale, or fur texture Step 4: Choose an appropriate background or environment for the animal you chose in Step 1 Step 5: Choose a real or fantasy scenario randomly that matches the animal you chose in Step 1 Step 6: Choose random lighting conditions and colors for the animal, describing how the chosen lighting affects the shadows and visual appearance of the scene Step 7: Begin the description with the following text: "/imagine prompt: An award-winning photograph of a [animal] in the [era] with [scenario] in [lighting conditions] and [color], featuring " Always begin with that phrase. Replace the information in [brackets] with your selection and then continue the description sensibly or as instructed. Don't forget the "/" at the very beginning. Step 8: Generate a unique, interesting, and concise (no more than 150 words) description of the image incorporating elements from Steps 1-6 Step 9: Use common photography composition and post-processing techniques and effects to enhance the quality of the photograph and create visual appeal. Add statement emphasizing the high-quality and visually stunning nature of the final image, using language that is easy to understand. Step 10: Always end prompt with the following text: “insanely high resolution, realistic, 8k, HD, HDR, XDR, focus + sharpen + wide-angle 8K resolution + photorealistic + film grain + HDR10 Ken Burns effect + Adobe Lightroom + rule-of-thirds + octane render + anamorphic + bokeh --ar 16:9 --s 1000 --q 2 --c 70” Make sure that this is always at the very end of the prompt!
NatureIsFire prompt script (Tango’one)
Prompting style called "NatureIsFire": Objective: Create a masterpiece photograph showcasing the wonder and beauty of the natural landscape, suitable for a gallery showcase. This script will generate prompt that will be used as an input into a generative AI art program, Midjourney. Several steps in the script include examples. In each case, do not be limited by the examples provided, use your imagination and judgement based on your contextual understanding of the concepts. Step 0: While executing this script, do not display interim outputs from each step. Only display the finalized resulting prompt. Step 1: Biome - Choose a biome from the following list: tropical rain forest, tropical dry forest, tropical savanna, desert, temperate grassland, temperate woodland and shrubland, temperate forest, northwestern coniferous forest, boreal forest, bamboo forest, tundra, arctic, pond, lake, shallow sea, deep sea, coral reef, iceberg, Pandora (from the movie Avatar). Step 2: Flora - Choose several types of plants, trees and other flora that would be present in the biome selected in step 1. Here are some examples for each biome, but do not be limited by these examples nor assume that you must include these specific examples. Use your best judgement to choose visually exciting elements for the scene. : Tropical rainforest: towering kapok trees, hanging orchids, delicate ferns, and colorful bromeliads Tropical dry forest: gnarled baobab trees, spiny cacti, and prickly pear Tropical savanna: tall grasses, acacia trees, and baobabs Desert: cacti, succulents, and sagebrush Temperate grassland: tall grasses, wildflowers, and prairie shrubs Temperate woodland and shrubland: oak trees, dogwoods, and flowering shrubs Temperate forest: maple trees, birch trees, and wildflowers Northwestern coniferous forest: towering redwoods, Douglas firs, and ferns Boreal forest: spruce trees, pine trees, and lichens Bamboo forest: towering bamboo stalks, lush ferns, and delicate orchids Tundra: low shrubs, mosses, and lichens Arctic: hardy grasses, wildflowers, and shrubs Pond: water lilies, lotus flowers, and cattails Lake: willow trees, reeds, and water hyacinths Shallow sea: vibrant coral, seaweed, and starfish Deep sea: bioluminescent creatures, colorful fish, and coral reefs Coral reef: vibrant coral, sea anemones, and tropical fish Iceberg: lichens, mosses, and algae Pandora: bioluminescent flora, towering trees, and exotic flowers Optional Step 3: Animals - Choose several types of animals that would be likely to be present in the selected biome. For example, if you chose a tropical rainforest, you might select toucans, jaguars, and monkeys. Step 4: Focal Feature - Choose a major natural feature to be used as the focal point of the image. For example, a misty mountain peak, a rushing river, or a colorful coral reef. Step 5: Time of Day - Choose a time of day for the image, such as sunrise, midday, or sunset. Step 6: Weather Event - Choose a weather type or weather event for the scene, such as a gentle rain, a thunderstorm, or a clear sky with fluffy clouds. Step 7: Use your imagination and creativity to come up with unique and interesting descriptions for our developing scene incorporating the selections in steps 1 through 6. Be sure to describe: the Flora including how the different types are placed in relation to each other, where the Animals (if selected) are located within the scene and what they are doing, what is unique or significant about the Focal Feature, how the weather event and time of day impact the lighting, shadows, colors and overall composition of the scene, how any Animals in the scene are interacting with each other or the Flora or the Focal Feature. Step 8: Create and include witty social media captions for the portrait, using relevant emojis, hashtags and clever word choices. Enclose the captions in a quote block within the prompt. For example: "Here's a snapshot of paradise 🌴☀️ #IslandLife #NatureIsFire-VI" Step 9: Specify the high-end camera model, professional lens, viewing angle and technical settings (e.g. focal length, timing, aperture, ISO) as well as any color, light or image filters to be applied. Step 10: Add the following phrase to the end of the prompt: " #NatureIsFire-VI + masterful composition + ultrasharp 8K resolution " Step 11: If there is Flora present in the scene, then append "+ high-detailed leaves + high-detailed bark " after Step 10 Step 12: If there is Animal(s) present in the scene, determine what type of outer layer they have. As appropriate based on that determination, append "+ high-detailed fur " "+ high-detailed scales "+ high-detailed feathers " after Step 11. Step 13: Select up to three additional visual effects to add to the portrait from the following list: wide-angle, god-rays, lens flare, Ken Burns effect, HDR10, ray-tracing, Behance, Adobe Lightroom, Unsplash, cinematography, epic and dramatic, rule-of-thirds, golden-ratio, #aweinspiring, #awardwinning, #masterpiece. Step 14: Start the prompt with: "/imagine prompt: RAW photo, an award winning HD photograph by National Geographic featuring ". Step 15: Close the prompt with " --ar 2:1 --s 1000 --v 4". Make sure this phrase is the last element in the prompt. Please generate a prompt now.