process of 5 "foolproof" prompts for the creator

does this sound familiar?

you sit in front of a blank screen, the cursor blinking in the input field as if tapping out the rhythm of your creative block.

in neighboring tabs, there are dozens of references, Google, tutorials; in your head, there's a crystal-clear, beautiful image that you've already conceived.

you take a breath and start typing: "girl, beautiful, holding a jar of cream, studio lighting, photorealism..."

enter...

the neural network thinks for a few seconds and produces... this.

a plastic face, six fingers on a hand, a crooked jar, overexposure, no volume in the picture; it's clear that it's mediocre AI content — a result you definitely can't show a client.

you try again, swap words, add "4k, octane render, cinematic." the result is slightly better but still soulless and formulaic.

dozens of generations and wasted credits later, your self-confidence melts away. you look at the work of other AI creators and feel that unpleasant sting of envy and despair.

"why does it work for them but look like nonsense for me?"

this frustration is the sharpest pain for any creator trying to make friends with artificial intelligence, which i've told you more than once. but all that aside...

here is what i want to tell you: it's not about you

a neural network is a set of perfectly fitted parts, and a prompt is not magic; it's the assembly instructions. today, i'll give you my versions of the instructions and the process.

this guide isn't just a collection of tips; it's a foolproof creative process that will reveal how else you can prompt and generate to improve the quality of your work.

ready?

so, here is our plan

before we dive into the details, let's look at our journey from a bird's-eye view. i've tried to outline the workflow with generations to be more or less universal, so it can be used for both video and static images.

the whole process, from a blurry idea in your head to a finished short animation (or just a series of frames), can be represented by five logical steps. each step is one prompt that solves one specific task and gives you an instant "quick win."

this workflow is your insurance against failure; it forces you to first think like a strategist who approaches the process sequentially.

now let's break down each step and each prompt in detail.

step 1. "Visionary Blueprint"

(prompt for idea development)

most failures begin not with a bad prompt, but with the lack of a clear idea. we rush to generate without answering the main questions: what do i want to say? to whom? why? how exactly?

this prompt forces you to stop and design the "soul" of your future project. it turns ChatGPT (or another text AI) into your personal creative director who helps crystallize the idea.

1. copy the prompt text below
2. paste it into a chat, for example, with ChatGPT
3. fill in the section and gaps as detailed as possible; the more context you provide, the deeper the result will be
4. send it and get a structured creative brief for your project

role:
you are a creative concept director and brand strategist specializing in visual communications.

role characteristics:
your thinking style is a fusion of marketer analytics and art director intuition. you don't just generate ideas; you design the viewer's emotional experience. your main superpower is the ability to turn blurry goals and abstract concepts into a clear, structured, and inspiring creative brief. you ask deep questions to uncover the essence of the project.

task:
based on the provided project data, you must create a detailed set of ideas for a series of visual generations (images and/or videos). this set should become the foundation for all further work with generations and include all the necessary information to create a creative idea.

project data:
1. product/topic: [enter data]
2. target audience: [enter data]
3. key content message: [e.g.: our product is not just hydration, but a real treat for the body]
4. main content task: [e.g.: to evoke an irresistible desire to try the product, convey its texture and scent through the picture, create a feeling of luxurious but natural care]
5. where it will be used, what format: [e.g.: Reels, 9:16, posts in a Telegram channel]

steps to follow:
1. analyze the key message and break it down into visual and emotional "atoms." find the main metaphor. use the visual metaphor method; pay attention to interesting 2nd and 3rd level metaphors. avoid cliché metaphors. form 10 ideas for a future promotional video or a series of static images from these metaphors.
2. deepen the idea: analyze the found metaphors and ideas. analyze and deepen all found ideas; make them creative, modern, and relevant for 2025.
3. determine "visual temperature": define the overall mood and atmosphere of each idea; describe it in 2-3 adjectives.
4. development of palette and light: suggest 2-3 color palette options (3-4 key colors) and describe the lighting scheme for each idea.
5. selection of key images and textures: list 5-7 key objects, textures, and symbols that should appear in the frame. think associatively; perform this step for each idea.
6. formulation of the "emotional anchor": describe in one sentence the main feeling the viewer should experience after watching.

answer:
present the result as a clearly structured detailed set of 10 ideas, each including:
1. main metaphor:
2. visual temperature:
3. color palette and light:
4. key images and textures:
5. emotional anchor
6. final concept (summary): (combine all points into one paragraph describing the idea)

step 2. "Character Sculptor"

(prompt for creating a portrait based on a moodboard)

i assume you've already found and chosen an idea. now you need a main character. and let's be honest, the hardest task is explaining to the neural network exactly how they should look. you need a detailed portrait from a collection of references.

this prompt turns AI into an analyzer of your moodboard (a set of reference images) and "translates" it into a language neural networks understand.

1. assemble a moodboard: find images that convey the type, style, and hairstyle you need
2. copy the prompt below
3. paste it into a chat with a neural network that can analyze images
4. get a comprehensive text description of the character, ready for the next step

role:
you are a hybrid of an art director and a concept artist, specializing in creating photorealistic characters.

role characteristics:
you are able to not just describe pictures but synthesize them into a single, cohesive, and detailed image. you can see the overall aesthetic behind individual features, read the mood, and translate the visual language of the moodboard into precise text descriptors for text-to-image neural networks.

task:
analyze the provided moodboard of several reference images and create based on them one comprehensive prompt for generating a character image. this prompt should be composed according to the correct formula so that it can be used to generate consistent images in Midjourney or other similar models.

steps to follow:
1. appearance synthesis: analyze all images and create an averaged but detailed portrait-description of the appearance. pay attention to face shape, eye color, eyebrow shape, lips, skin tone, presence of freckles or other features.
2. hairstyle and style description: describe the hairstyle, color, length, and texture of the hair; determine the overall clothing style seen in the references.
3. emotional state analysis: the emotion on the face should always be neutral.
4. determination of the lighting scheme and photo characteristics: it should always be a neutral full-face portrait shot on a white background.

answer:
present the result as a structured prompt using the formula: [main object + adjectives + action], [in environment / place / context], [surrounded by secondary elements / scene details], [with lighting / color atmosphere / effects], [in the style of...], [emotional message or mood].
use only the English language.
main object — describe the character with details of appearance, hairstyle, clothing style; the emotion on the face should always be neutral. it should always be a neutral full-face portrait shot on a white background. focus on a detailed description of the character itself.

example

screenshot of the Pinterest moodboard i assembled:

generations based on the AI prompt:

step 3. "First Frame"

(prompt for generating key visuals)

turning AI into your personal commercial photographer who understands the task precisely.

1. take the English description of the character you received in step 2
2. take the idea from step 1
3. formulate a brief description of your product (or ask AI to do it)
4. take the prompt structure below, complete it with missing information about the character, product, environment, and events based on your idea
5. paste the prompt into the neural network
6. press *enter* and prepare to be amazed

a hyper-detailed, photorealistic medium shot of [insert full character description from step 2], she is gently holding/wearing [select and insert your product description here], presenting it towards the camera, in a beautifully styled [environment], the background is slightly blurred, focusing on the character and the product, with [lighting], in the style of [style description + parameters]

my example:

character:

product:

generation:

step 4. "Consistency Engine"

(prompt for creating additional angles)

one perfect frame is great, but for storytelling or video storyboarding, you need at least a few. and here arises the main headache for AI creators — consistency.

this step solves exactly this task. it will create different variations of your current prompt. there are two ways: you can give AI simple prompts with small changes, or use my structure and scheme to generate additional prompts to get a significantly more interesting and cooler result.

1. select the best image generated in step 3
2. copy the prompt below
3. send the prompt to a text AI and get new ones for generating frames

here is what i got:

role:
you are a hybrid of an art director and a concept artist, specializing in creating photorealistic images. your task is to adapt a user-provided prompt for new angles, emotions, and variations of the same scene. you can see the overall aesthetic behind individual features, read the mood, and translate visual language into precise text descriptors for text-to-image neural networks.

you need to analyze the provided prompt and image, then create 10 variations of what is happening in this scene by replacing descriptions of camera angles, hero emotions, and hero actions. at the same time, the description of the character, environment, and object in the frame must not change.

general prompt formula: [main object + adjectives + action], [in environment / place / context], [surrounded by secondary elements / scene details], [with lighting / color atmosphere / effects], [in the style of...], [emotional message or mood].

starting prompt: [insert yours]

response format: 10 full texts of prompts modified according to instructions with different variations of the same shot, with different angles and emotions, but maintaining the overall picture.

step 5. "Animator"

(prompt for creating animation)

you have a series of stunning static frames, and the final touch is to add movement. but it's not always clear how to do this.

this structure is intended for text-to-video neural networks. it teaches you to describe not just the image, but the action and camera movement.

1. select frames for animation
2. upload it to the chosen video neural network as a source image (or take two as start and end)
3. take the structure below
4. adapt the description to your scene
5. run the generation and get a short animated video

[stylistic medium (camera or film type)][narrative core][main action (protagonist, actor)][environmental context][atmospheric details][shot and angle][lens and focus][lighting and color][camera movement]

then use this prompt in the neural network for animation, edit if necessary, and get the finished project.