Prompt Generator

Build a fully functional, responsive, one-page web application called “AI Prompt Studio.”
VERSION 1 SCOPE
This first version must generate prompts for AI VIDEO models only.
Do not add the following features yet:
● Do not add a random idea generator.
● Do not add a guided questionnaire.
● Do not add an image prompt generator.
● Do not add prompt history.
● Do not add a template library.
● Do not add user profiles.
● Do not add payments, pricing, subscriptions, or monetization features.
● Do not add community features.
● Do not add analytics dashboards.
● Do not add a marketing landing page, testimonials, blog, or unnecessary navigation.
Those features may be added later through separate editing prompts.
CORE PURPOSE
The application should help content creators transform one simple video idea into three
different production-ready prompts:
1. A prompt optimized for a simple AI video model.
2. A prompt optimized for a medium-capability AI video model.
3. A prompt optimized for a powerful AI video model.
The three results must not simply be the same prompt with different lengths. Each version
must genuinely adapt its complexity, number of actions, camera direction, scene structure,
and production detail to what that level of model can realistically understand.
The user should be able to copy any result and paste it directly into an AI video generator
without editing it first.
This must be a real, functional AI application, not a static mockup.
Use Base44’s native AI or agent capabilities to generate the prompts dynamically. Do not
require an external API key. Do not use hardcoded prompt results or rotate between a fixed
list of examples.
The AI generation should happen behind the interface. Do not expose a chat-style interface
to the user.
APPLICATION LANGUAGE
Build the complete user interface in natural English.
The generated prompts must also be written in natural English.
The user may enter their initial idea in English, Spanish, or another language. Detect the
meaning of the input and silently produce the final prompts in English without mentioning
that a translation took place.
PAGE STRUCTURE
Create one focused application page.
Do not create a large landing page before the tool.
The generator itself should be visible immediately when the page opens.
Use this general structure:
1. Compact header.
2. Main generator workspace.
3. Input and settings panel.
4. Generated results panel.
HEADER
Display:
Title: AI Prompt Studio
Subtitle: Turn one simple idea into a production-ready AI video prompt.
You may include a small, simple video or spark icon beside the title, but do not use an
oversized logo or decorative hero section.
MAIN DESKTOP LAYOUT
Use a balanced two-column layout:
Left column:
● Idea input.
● Generation settings.
● Main generate button.
Right column:
● Empty-state message before generation.
● Three result cards after generation.
On mobile, stack the left column above the results.
Keep the generator centered and readable on normal laptop and desktop screens used for
screen recordings.
IDEA INPUT
Add a large textarea with this label:
Describe your video idea
Use this placeholder:
A medieval knight in red armor discovers a dragon inside a frozen castle...
The placeholder is only an example. Do not automatically insert it as actual input.
The textarea should accept anything from a very short concept to a more complete
description.
Do not require the user to write camera instructions, lighting, transitions, or technical details.
The application should intelligently develop those elements from the basic idea.
SETTINGS
Place the settings directly below the idea textarea.
Setting 1:
Label: Maximum prompt length
Use a numeric field and a matching slider.
The value represents characters, including spaces.
Configuration:
● Minimum: 250 characters.
● Maximum: 5,000 characters.
● Default: 1,800 characters.
● Step: 50 characters.
Display the selected number clearly.
This maximum applies separately to every generated prompt.
No result may exceed the selected maximum character count.
Do not simply cut the text in the middle of a sentence. If a generated result is too long,
automatically rewrite or condense it until it fits naturally within the limit.
The prompts may be shorter than the maximum when additional text would only add
repetition or unnecessary detail.
Setting 2:
Label: Video orientation
Create a clear segmented selector with exactly these two options:
● Horizontal — 16:9
● Vertical — 9:16
Default: Horizontal — 16:9
Do not add square or other aspect ratios in this first version.
Setting 3:
Label: Visual style
Create a dropdown menu containing:
● Auto
● Cinematic
● Photorealistic
● Commercial
● Documentary
● Fantasy
● Science Fiction
● Horror
● Anime
● 3D Animation
● Vintage Film
● Product Advertisement
Default: Auto
When Auto is selected, infer the most suitable visual treatment from the idea.
When another option is selected, apply that style consistently without overriding the central
subject, action, or story supplied by the user.
Do not imitate a specific living artist.
MAIN ACTION
Add one prominent primary button:
Generate Prompts
The button should remain disabled until the idea field contains meaningful text.
When generation begins:
● Disable the button temporarily.
● Preserve all user inputs and settings.
● Display a polished loading state.
● Use a short loading message such as “Building your prompts...”
● Prevent duplicate generation requests caused by repeated clicks.
Do not clear the idea or settings after generation.
EMPTY RESULTS STATE
Before the first generation, show a clean empty-state panel on the right.
Use this message:
Your optimized prompts will appear here.
Add a short supporting line:
You’ll get a different version for simple, medium, and powerful video models.
Do not display fake example outputs in the results area.
RESULTS
After generation, replace the empty state with three separate result cards in this exact order:
1. Simple Model
2. Medium Model
3. Powerful Model
Each card must contain:
● The model-level title.
● A short one-line explanation.
● The generated prompt.
● A live character count.
● A Copy Prompt button.
Use these explanations:
Simple Model: Clear, focused, and easy for less advanced generators to follow.
Medium Model: More detailed motion, camera direction, atmosphere, and visual control.
Powerful Model: A fully developed prompt for advanced models that can follow complex
direction.
The prompt itself must be the main visual focus of each card.
The result text must remain selectable.
COPY FUNCTION
Each card must have its own button labeled:
Copy Prompt
When clicked:
● Copy only the generated prompt.
● Do not copy the card title, explanation, quotation marks, or character count.
● Briefly change the button state to “Copied.”
● Return it to “Copy Prompt” after a short delay.
The copy function must work on desktop and mobile.
AI GENERATION ARCHITECTURE
Create a dedicated internal AI generation action or Base44 agent named:
Prompt Director
Connect the form directly to this AI action.
The user should never have to open or interact with an AI chat.
The generation action receives:
● The original idea.
● The selected maximum character count.
● The selected orientation.
● The selected visual style.
It returns exactly three fields:
● simple
● medium
● powerful
Use a reliable structured response similar to:
{
"simple": "Generated prompt text",
"medium": "Generated prompt text",
"powerful": "Generated prompt text"
}
Do not display the JSON structure to the user.
Parse it and place each value in the correct result card.
If the AI produces invalid structured data, automatically retry or safely recover the three
prompt strings instead of showing broken JSON to the user.
PROMPT DIRECTOR — CORE ROLE
Configure Prompt Director as an expert AI video prompt designer, cinematic director,
storyboard planner, and prompt editor.
Its purpose is to turn a basic creative idea into clear, model-appropriate instructions that
improve visual coherence, action, camera work, continuity, and prompt adherence.
It must internally analyze:
● The main subject.
● Important identifying details.
● The setting.
● The central action.
● The chronological progression.
● The visual style.
● The intended orientation.
● The composition.
● The camera position.
● Camera movement.
● Subject movement.
● Environmental movement.
● Lighting.
● Atmosphere.
● Color treatment.
● Physical continuity.
● Character continuity.
● Object continuity.
● Dialogue or audio only when requested.
● The amount of complexity suitable for each model level.
Do not show this analysis to the user.
Return only the finished prompts.
GENERAL PROMPT-WRITING RULES
Every generated prompt must:
● Preserve the user’s central idea.
● Be self-contained.
● Be ready to paste directly into an AI video generator.
● Use natural, descriptive English.
● Describe events in a clear chronological order.
● Make the main subject immediately understandable.
● Make actions visually observable.
● Use purposeful camera direction.
● Keep movement physically coherent.
● Maintain the appearance and identity of recurring subjects.
● Maintain the position and identity of important objects.
● Respect the selected horizontal or vertical orientation.
● Apply the selected visual style consistently.
● Remain within the selected maximum character count.
● Avoid contradictions.
● Avoid redundant adjectives.
● Avoid generic keyword spam.
● Avoid unrelated additions.
● Avoid introducing extra characters that the user did not need.
● Avoid cramming too many events into one short clip.
● Avoid abrupt teleportation, unexplained transformations, or impossible continuity
unless the user asks for them.
● Avoid real brands, logos, celebrities, copyrighted characters, or recognizable
franchises unless the user explicitly includes them.
● Avoid defaulting to cliché neon cyberpunk cities or generic office scenes.
● Avoid unrequested subtitles, captions, interface elements, logos, and watermarks.
● Avoid model-specific parameter syntax unless the user explicitly provides it.
● Avoid adding resolution, frame rate, seed values, or technical commands that were
not requested.
● Avoid mentioning the words “simple model,” “medium model,” or “powerful model”
inside the generated prompt.
● Avoid introductions such as “Here is your prompt.”
● Avoid explanations after the prompt.
● Avoid quotation marks around the entire prompt.
● Avoid Markdown headings or bullet points unless a powerful multi-shot prompt
genuinely benefits from short shot labels.
Do not write a negative-prompt section separately.
Express important visual constraints naturally inside the main prompt.
If the user’s idea is vague, make intelligent creative decisions, but do not replace the idea
with a different concept.
If the user gives specific details, preserve them accurately.
If the idea contains several characters, distinguish them clearly by appearance, clothing,
position, or role so the model does not confuse them.
If the idea includes a recurring character across different shots, repeat only the essential
identity details needed to maintain consistency.
If the user includes exact dialogue:
● Preserve the exact spoken words.
● Place the dialogue inside quotation marks.
● Identify who says each line.
● Keep the amount of dialogue realistic for the scene.
● Do not rewrite the dialogue unless the user asks.
● Do not add subtitles unless requested.
If the user does not request dialogue, do not invent spoken lines.
If the user requests sound or audio, include useful sound direction.
If the user does not mention audio, prioritize visual instructions and do not overload the
prompt with unnecessary sound design.
Do not make every result a static shot.
Use purposeful movement when the idea supports it, while keeping the amount of movement
appropriate for the selected model level.
SIMPLE MODEL PROMPT RULES
The Simple Model result should:
● Focus on one central visual idea.
● Prefer one continuous shot.
● Use one clear location.
● Use one main action.
● Keep the number of characters limited.
● Use one understandable camera angle or one simple camera movement.
● Avoid complicated transitions.
● Avoid rapid scene changes.
● Avoid several simultaneous actions.
● Avoid complex continuity requirements.
● Use straightforward language.
● Prioritize subject, action, setting, camera, lighting, style, and orientation.
It should usually use only part of the available character allowance.
Do not make it long merely because a high maximum was selected.
MEDIUM MODEL PROMPT RULES
The Medium Model result should:
● Add richer environmental detail.
● Include a clear beginning and development of the action.
● Support one or two connected action beats.
● Use controlled camera movement.
● Add lighting and atmospheric direction.
● Clarify subject movement and environmental response.
● Maintain one coherent scene or a very simple transition.
● Include stronger composition and visual hierarchy.
● Remain easy to follow.
● Avoid excessive shot changes.
It should contain materially more control than the Simple Model version, not just more
adjectives.
POWERFUL MODEL PROMPT RULES
The Powerful Model result should:
● Use the full creative potential of an advanced AI video model.
● Develop a precise visual progression from beginning to end.
● Include sophisticated but coherent camera direction.
● Include multiple connected shots only when the idea benefits from them.
● Keep every shot logically connected.
● Preserve subject and environment continuity.
● Describe important transitions clearly.
● Coordinate camera movement, subject action, background movement, and lighting.
● Include cinematic pacing.
● Include audio or dialogue direction only when relevant.
● Make the sequence dynamic and visually engaging.
● Avoid overcrowding the clip with more actions than can be represented coherently.
● Prioritize clarity over decorative language.
When using multiple shots, normally limit the sequence to a small number of meaningful
shots unless the user specifically requests a longer or more complex sequence.
Do not turn the Powerful Model prompt into an unstructured wall of cinematic adjectives.
MODEL-LEVEL DIFFERENTIATION
The three results must clearly differ in production complexity:
Simple: One direct visual concept with minimal dependencies.
Medium: A richer version with controlled movement, atmosphere, and a more developed
action.
Powerful: A fully directed sequence with advanced temporal, cinematic, and continuity
control.
Do not make the Simple prompt merely a truncated version of the Powerful prompt.
Do not make the Powerful prompt merely a longer version of the Simple prompt.
Each one should be independently written for its intended model capability.
CHARACTER-LIMIT HANDLING
Before displaying each result:
1. Count all characters, including spaces.
2. Confirm the result does not exceed the selected maximum.
3. If it exceeds the maximum, rewrite it naturally.
4. Recount it.
5. Display it only when it fits.
Never cut a sentence halfway.
Never remove the central subject, action, orientation, or selected visual style merely to meet
the character limit.
When the limit is unusually low, prioritize:
1. Main subject.
2. Main action.
3. Setting.
4. Camera.
5. Visual style.
6. Orientation.
7. Essential continuity.
ERROR HANDLING
If the idea is empty, show:
Enter a video idea before generating prompts.
If generation fails, show a clear inline message:
The prompts could not be generated. Please try again.
Include a Retry button.
Do not erase the user’s idea or selected settings after an error.
Do not expose technical stack traces, raw API errors, or invalid JSON.
DESIGN DIRECTION
Create a clean, premium, creator-focused interface.
Use:
● A dark neutral background.
● High-contrast text.
● Restrained accent colors.
● Clear visual hierarchy.
● Rounded but professional cards.
● Generous spacing.
● Subtle borders.
● Soft shadows.
● Clean typography.
● Small, purposeful transitions.
● Easily readable prompt text.
● Clearly differentiated result-card badges.
Avoid:
● Bright neon visual effects.
● Excessive gradients.
● Animated backgrounds.
● Decorative 3D objects.
● Large marketing banners.
● Clutter.
● Tiny controls.
● Excessive glassmorphism.
● Distracting animations.
The interface should look polished in a YouTube screen recording.
RESPONSIVE BEHAVIOR
Desktop:
● Use two columns.
● Keep the input panel visible beside the results.
● Allow long prompt text to wrap naturally.
Tablet:
● Use a balanced layout or stack if necessary.
Mobile:
● Stack the input panel above the result cards.
● Make buttons full-width where appropriate.
● Keep settings easy to tap.
● Prevent horizontal scrolling.
● Ensure prompt cards remain readable.
ACCESSIBILITY
● Use proper labels for every field.
● Ensure keyboard navigation works.
● Ensure visible focus states.
● Use accessible contrast.
● Do not rely only on color to identify the three model levels.
● Add accessible names to icon buttons.
● Make copy confirmations understandable to screen readers.
DATA AND PRIVACY
Do not create a persistent database for this first version.
Do not save user ideas or generated prompts permanently.
Keep the current idea, settings, and results in the current application session only.
Do not request personal information.
Do not create an app-level login or onboarding flow.
FINAL ACCEPTANCE CRITERIA
The first build is complete only when all of the following are true:
1. The application opens directly on the video prompt generator.
2. The user can enter one simple idea.
3. The user can set a maximum character count.
4. The user can choose horizontal or vertical video.
5. The user can select a visual style.
6. The Generate Prompts button calls a real AI generation action.
7. Three genuinely different prompts are returned.
8. The three outputs correspond to Simple, Medium, and Powerful models.
9. Every result respects the selected character maximum.
10. Every result reflects the selected orientation.
11. Every result reflects the selected visual style.
12. Every result can be copied independently.
13. No random generator exists yet.
14. No questionnaire exists yet.
15. No image prompt mode exists yet.
16. There are no hardcoded or mock prompt results.
17. The interface works on desktop and mobile.
18. No unnecessary page, dashboard, pricing block, or marketing section has been
added.
Before considering the build complete, inspect the generated application and fix any
functional error you find.
Do not stop after creating the visual interface.
Make the complete generation flow work.

Modify the existing AI Prompt Studio application. Do not rebuild the application from scratch.
Preserve all existing functionality, visual styling, labels, settings, AI generation rules, result
cards, character-limit handling, responsive behavior, and copy buttons.
Add one new feature only:
A random video idea generator for users who do not know what to create.
PLACEMENT
Add a secondary button close to the “Describe your video idea” textarea.
Label the button:
Random Idea
Use a small dice or spark icon if appropriate.
The button must be visually secondary to the main “Generate Prompts” button.
BEHAVIOR
When the user clicks Random Idea:
1. Use the application’s real AI capability to generate one fresh video concept.
2. Place that concept inside the existing idea textarea.
3. Do not generate the three final prompts automatically.
4. Let the user review or edit the random idea first.
5. Keep the current maximum character count, orientation, and visual style unchanged.
6. The user should still press Generate Prompts when ready.
The generated idea should be concise enough to edit easily.
Use approximately one to three natural sentences.
It should describe:
● A clear main subject.
● An interesting setting.
● A visually understandable action.
● A useful creative direction or mood.
It should be an idea, not a fully developed production prompt.
Do not fill it with camera instructions, technical parameters, or long shot descriptions.
VARIETY RULES
Generate varied concepts across areas such as:
● Cinematic action.
● Nature.
● Adventure.
● Historical scenes.
● Science fiction.
● Fantasy.
● Suspense.
● Emotional storytelling.
● Commercial concepts.
● Product visuals.
● Surreal but coherent ideas.
● Documentary-style moments.
Do not default repeatedly to:
● Neon cyberpunk cities.
● Generic people typing in offices.
● Generic robots walking through cities.
● Empty landscape shots with no clear action.
Do not introduce real brands, celebrities, copyrighted characters, or recognizable franchises.
The random idea may use the currently selected visual style as inspiration, but it must
remain editable and should not repeat the setting label mechanically.
If the selected orientation is vertical, favor ideas that can compose well vertically.
If the selected orientation is horizontal, favor ideas that can use a wider environment or
cinematic composition.
REPETITION CONTROL
Within the current session, avoid returning the same or nearly identical idea repeatedly.
If possible, keep a temporary session-only list of recent random ideas and instruct the AI to
generate something meaningfully different.
Do not create permanent storage or a database.
LOADING AND ERROR STATES
While generating:
● Disable the Random Idea button.
● Display “Creating an idea...”
● Do not remove the user’s current idea until the new one has been generated
successfully.
If generation fails:
● Keep the previous idea.
● Return the button to its normal state.
● Display a small error message and allow another attempt.
Do not change any other part of the application.
After implementing this feature, test that Random Idea fills the textarea, does not generate
the final prompts automatically, and does not alter the existing settings.

Modify the existing AI Prompt Studio application. Do not rebuild it from scratch.
Preserve:
● The existing quick idea workflow.
● The Random Idea button.
● Maximum prompt length.
● Video orientation.
● Visual style.
● The three Simple, Medium, and Powerful prompt results.
● All existing Prompt Director rules.
● Copy buttons.
● Loading states.
● Error handling.
● Responsive design.
● Current visual styling.
Add a second way to create a VIDEO prompt:
A short guided questionnaire that asks the user what kind of video they want to make and
produces one final prompt based on their answers.
MODE SELECTION
At the top of the input panel, add two clearly visible modes:
● Quick Idea
● Guided Builder
Default: Quick Idea
Quick Idea must preserve the existing textarea, Random Idea button, settings, and
three-result workflow exactly as they already work.
Do not remove or redesign that workflow unnecessarily.
GUIDED BUILDER STRUCTURE
When Guided Builder is selected, replace the large idea textarea area with a short
step-by-step questionnaire.
Keep the existing global settings visible:
● Maximum prompt length.
● Video orientation.
● Visual style.
Also add a target model selector for Guided Builder with exactly these options:
● Simple Model
● Medium Model
● Powerful Model
Default: Powerful Model
The questionnaire should feel short and easy to complete.
Use a compact wizard with progress such as:
Question 1 of 6
Include Back and Next buttons.
Do not open a new page.
Keep the questionnaire inside the existing generator workspace.
QUESTIONS
Question 1:
Label: What is the main subject?
Supporting text: Describe the main person, creature, object, product, or place.
Example placeholder: A medieval knight wearing detailed red armor
Required.
Question 2:
Label: What should happen?
Supporting text: Describe the main action or the progression you want to see.
Example placeholder: He enters a frozen castle and discovers a dragon awakening beneath
the ice
Required.
Question 3:
Label: Where and when does it happen?
Supporting text: Describe the location, time period, weather, or time of day.
Example placeholder: Inside an ancient frozen castle during a snowstorm at night
Optional, but encourage an answer.
Question 4:
Label: What mood should the video have?
Use selectable options:
● Auto
● Epic
● Mysterious
● Emotional
● Suspenseful
● Energetic
● Peaceful
● Dark
● Playful
● Commercial
Allow only one selection.
Default: Auto
Question 5:
Label: How should the camera and movement feel?
Use selectable options:
● Auto
● Slow and cinematic
● Dynamic and energetic
● Smooth tracking shot
● Handheld and realistic
● Dramatic camera changes
● Mostly static and controlled
Default: Auto
Also allow a short optional custom note.
Question 6:
Label: Should it include dialogue or specific sound?
Use a multiline optional field.
Supporting text: Leave this empty if you only want visual direction. Put exact dialogue inside
quotation marks.
Example placeholder: The knight whispers, “Something is moving under the ice.” Add
cracking ice and distant dragon sounds.
FINAL GUIDED ACTION
On the final step, display a button:
Create My Prompt
When clicked:
1. Combine the questionnaire answers with the selected orientation, visual style,
maximum character count, and target model level.
2. Send them to the existing Prompt Director AI system.
3. Generate one final video prompt.
4. Apply the same quality, continuity, character-limit, dialogue, camera, style, and
orientation rules already used by the application.
5. Adapt complexity to the selected target model level.
GUIDED RESULT
Guided Builder should return one result card rather than three.
Title:
Your Guided Prompt
Display a small badge showing the selected target model level.
Include:
● The completed prompt.
● Live character count.
● Copy Prompt button.
● Start Over button.
Copy Prompt must copy only the prompt.
Start Over should clear the questionnaire answers after confirmation but preserve the global
settings unless the user chooses to reset them manually.
Do not display a questionnaire summary instead of the prompt.
Do not produce an explanation before or after the prompt.
MODEL-LEVEL BEHAVIOR
If Simple Model is selected:
● Create one focused and easy-to-follow scene.
● Reduce simultaneous actions.
● Avoid complicated transitions.
● Use simple camera direction.
If Medium Model is selected:
● Add richer motion, atmosphere, lighting, and one or two connected action beats.
● Keep the progression coherent.
If Powerful Model is selected:
● Create a more advanced and fully directed prompt.
● Use complex camera movement or multiple connected shots only when appropriate.
● Preserve continuity.
● Coordinate action, environment, camera, and sound.
STATE MANAGEMENT
Allow the user to switch between Quick Idea and Guided Builder without losing work during
the current session.
If the user returns to Quick Idea, preserve the existing idea and any generated quick results.
If the user returns to Guided Builder, preserve questionnaire progress and its generated
prompt until the page is refreshed or the user presses Start Over.
Do not create persistent storage or a database.
RESPONSIVE DESIGN
On mobile:
● Keep each question easy to read.
● Make Back, Next, and Create My Prompt easy to tap.
● Prevent the questionnaire from overflowing the screen.
● Keep the progress indicator visible.
FINAL CHECK
After implementing the feature, verify:
● Quick Idea still generates three results.
● Random Idea still works.
● Guided Builder asks six questions.
● Guided Builder produces one prompt.
● The chosen model strength changes the prompt complexity.
● The final prompt respects the character maximum.
● Orientation and visual style are included.
● Copy Prompt works.
● Switching modes does not destroy the other mode’s current work.

Modify the existing AI Prompt Studio application. Do not rebuild it from scratch.
Preserve every existing VIDEO feature exactly:
● Quick Idea mode.
● Guided Builder mode.
● Random Idea button.
● Maximum prompt length.
● Horizontal and vertical orientation.
● Visual style selector.
● Simple, Medium, and Powerful video results.
● Guided single-prompt result.
● Copy buttons.
● Session state.
● Loading states.
● Error handling.
● Responsive design.
● Prompt Director video rules.
Add a second top-level content type:
● Video Prompts
● Image Prompts
Default: Video Prompts
Place this selector near the top of the main generator workspace, below the application title
and above Quick Idea and Guided Builder.
The selector must make it obvious whether the user is currently creating a video prompt or
an image prompt.
Do not create a separate page.
Do not remove or alter the existing video generator.
UPDATE THE HEADER
Keep the application title:
AI Prompt Studio
Update the subtitle to:
Turn one simple idea into a production-ready AI video or image prompt.
IMAGE PROMPTS — QUICK IDEA MODE
When Image Prompts and Quick Idea are selected:
Keep the same general layout as the video generator.
Change the idea label to:
Describe your image idea
Use this placeholder:
A futuristic glass house built into a snowy mountain at sunrise...
Keep:
● Maximum prompt length.
● Orientation.
● Visual style.
● Random Idea.
● Generate Prompts.
Rename the main button in image mode:
Generate Image Prompts
The orientation options remain:
● Horizontal — 16:9
● Vertical — 9:16
Do not add square orientation unless requested later.
IMAGE RESULTS
In Quick Idea image mode, generate three result cards:
1. Simple Image Model
2. Medium Image Model
3. Powerful Image Model
Use these explanations:
Simple Image Model: A clear, focused prompt with the most important visual details.
Medium Image Model: More control over composition, lighting, environment, and
appearance.
Powerful Image Model: A highly developed prompt with advanced composition and precise
visual direction.
Each card must include:
● Generated prompt.
● Character count.
● Copy Prompt button.
The generated text must respect the same selected maximum character count.
IMAGE PROMPT AI LOGIC
Extend the existing Prompt Director system so it can distinguish clearly between video and
image generation.
When content type is Image Prompts, create still-image prompts.
Do not write video instructions.
Do not include:
● Camera movement.
● Shot transitions.
● Temporal progression.
● Scene cuts.
● Sound effects.
● Dialogue delivery.
● Lip sync.
● Animation instructions.
● “The camera moves.”
● “The scene transitions.”
● “The video begins.”
● “The clip ends.”
Instead, internally analyze:
● Main subject.
● Subject appearance.
● Pose or visible action.
● Environment.
● Composition.
● Framing.
● Camera angle.
● Lens characteristics when useful.
● Foreground, middle ground, and background.
● Lighting direction.
● Time of day.
● Color palette.
● Materials.
● Textures.
● Depth.
● Atmosphere.
● Visual style.
● Orientation.
● Important spatial relationships.
● Intended focal point.
● Level of detail suitable for the selected model strength.
Every image prompt must:
● Be self-contained.
● Be ready to paste into an AI image generator.
● Preserve the user’s central idea.
● Establish a clear focal subject.
● Use coherent composition.
● Respect the selected orientation.
● Apply the selected style.
● Avoid contradictory visual instructions.
● Avoid unnecessary keyword spam.
● Avoid unrelated characters or objects.
● Avoid real brands, celebrities, copyrighted characters, or living artists unless the user
explicitly requests them.
● Avoid accidental interface elements, watermarks, captions, and logos.
If the user explicitly requests visible text inside the image:
● Preserve the exact text inside quotation marks.
● Describe where it should appear.
● Describe its visual treatment only when useful.
● Do not change the wording.
If the user does not request text:
● Do not add signs, captions, letters, logos, or readable text.
● Phrase the prompt to discourage unintended visible text when appropriate.
SIMPLE IMAGE MODEL
The Simple Image Model result should:
● Focus on one main subject.
● Use a clear composition.
● Include the essential setting.
● Add basic lighting and style.
● Avoid too many secondary objects.
● Use straightforward language.
● Remain compact.
MEDIUM IMAGE MODEL
The Medium Image Model result should:
● Add richer environment and composition.
● Add more precise lighting.
● Add framing and camera-angle direction.
● Clarify textures, materials, atmosphere, and depth.
● Keep the scene coherent and easy to interpret.
POWERFUL IMAGE MODEL
The Powerful Image Model result should:
● Provide highly controlled composition.
● Establish foreground, middle ground, and background when useful.
● Use precise lighting and color direction.
● Describe important surfaces, materials, scale, atmosphere, and subject detail.
● Include lens or depth-of-field direction only when it improves the image.
● Preserve visual hierarchy.
● Remain coherent rather than becoming a list of quality keywords.
Do not simply add words such as “masterpiece,” “best quality,” or “8K” to make the powerful
version appear better.
IMAGE RANDOM IDEA
Make the existing Random Idea button context-aware.
When Video Prompts is selected:
● Continue generating video ideas exactly as before.
When Image Prompts is selected:
● Generate one fresh still-image concept.
● Fill the image idea field.
● Do not include a sequence of actions.
● Do not include sound.
● Do not generate the three final prompts automatically.
The random image idea should have:
● A clear subject.
● A visually interesting environment.
● A composition or visual concept.
● A useful mood or style direction.
IMAGE GUIDED BUILDER
When Image Prompts and Guided Builder are selected, adapt the questionnaire for still
images.
Use these questions:
Question 1: What is the main subject?
Question 2: What should the subject be doing, wearing, or displaying?
Question 3: Where should the image take place?
Question 4: What mood should the image have?
Question 5: How should the image be composed?
Composition options:
● Auto
● Close-up
● Portrait
● Medium shot
● Wide shot
● Low angle
● High angle
● Symmetrical
● Editorial
● Product-focused
Question 6: Are there any essential details or exact text that must appear?
Do not show the audio and dialogue question in image mode.
Keep the target model selector:
● Simple Model
● Medium Model
● Powerful Model
At the end, produce one image prompt labeled:
Your Guided Image Prompt
Include the selected model-level badge, character count, Copy Prompt, and Start Over.
CONTENT-TYPE STATE
Preserve video and image work separately during the current session.
Switching to Image Prompts must not erase the current video idea, video questionnaire
answers, or video results.
Switching back to Video Prompts must restore the previous video state.
Do not create permanent storage or a database.
DESIGN CONSISTENCY
Use the same design system for video and image modes.
Use a small video icon for Video Prompts and an image icon for Image Prompts if useful.
Clearly indicate the active mode without changing the entire layout.
Do not create two visually disconnected applications.
IMPORTANT FUNCTIONAL LIMIT
This application generates PROMPTS for image models.
It must not attempt to generate actual images.
Do not add an image gallery, image upload system, image editor, or image generation API.
FINAL CHECK
After implementation, test:
1. Video Prompts remains the default.
2. Existing video functionality still works.
3. Image Quick Idea produces three image prompts.
4. Image Guided Builder produces one final image prompt.
5. Image prompts contain no video motion, transitions, or audio language.
6. Random Idea adapts correctly to the selected content type.
7. Maximum character count works in image mode.
8. Horizontal and vertical orientation works in image mode.
9. Visual style works in image mode.
10. Copy buttons work in every result.
11. Switching modes preserves current session work.
12. No actual image generation was added.