Generate Custom Prompts for Missing ARASAAC Pictograms

[hectoritr]

Description
To address missing pictograms in ARASAAC, we need to create a specific prompt for each word that does not have an available pictogram. Each custom prompt should use the original prompt as contextual input to maintain consistency and accuracy in generating alternative image prompts.

Generate Custom Prompts:

For each missing pictogram, create a custom prompt that leverages the original prompt context.
The custom prompt should ensure the generated image closely matches the intended meaning and style of existing ARASAAC pictograms.

Prompt Structure:
Use the original prompt as context to construct each specific prompt.
Ensure each prompt is clear and detailed enough to generate images that align with ARASAAC's pictogram standards.

Acceptance Criteria

• Custom prompts are generated for each missing pictogram using the original prompt context.
• Prompts are ready to use for image generation to fill in gaps for words without pictograms.

[hectoritr]

@AliLee0923 please use this prompt, send it first to the Endpoint we already have (GPT3.5) and share the results with me, let's check if we need to upgrade the model for this and previous prompts.
Prompt

Help me write a prompt for an image generation model (Leonardo AI). Given a list of words I need to create a prompt for each one following the provided template. Change the <WORD> placeholder for a short description of the word and some context to create a precise image. 

Template: a pictogram of a <WORD> with minimal details, using bold lines and solid colors, white background. Use colors similar to the real ones.

List of words {Word1, Word2, Word3, WordN,...}

[hectoritr]

Let's add the original prompt from the User to add context to the Leonardo prompt.

[hectoritr]

New enhanced prompt.

Create a detailed prompt to generate a pictogram for '${word}'.
First, analyze the word using these criteria:

ACTIONS (Classification Criteria):
-Can it be performed/demonstrated?
-Does it involve movement or change?
-Can you complete the phrase "to [word]"?
-Examples: run, eat, jump, write, dance

OBJECTS (Classification Criteria):
-Can it be touched or physically exist?
-Is it a person, place, or thing?
-Can you put "the" or "a" before it?
-Examples: chair, dog, house, tree, book

ADJECTIVES (Classification Criteria):
-Does it describe a quality or state?
-Can you put "very" before it?
-Can you add "-er" or "-est" to compare it?
-Examples: tall, hot, happy, heavy, bright

Once classified, follow the appropriate template:
For ACTIONS:
-Show a simplified human figure mid-action
-Capture the most distinctive moment of the action
-Include motion lines or indicators where needed
-Use side view for directional actions (run, jump)
-Use front view for symmetrical actions (stretch, dance)
-Include only props essential to understand the action

For OBJECTS:
-Show the complete item in its most recognizable form
-Use the view that shows key identifying features
-For animals: side profile, all limbs visible
-For tools/items: standard usage orientation
-For buildings/structures: front-facing view
-Avoid showing interaction or movement

For ADJECTIVES:
-Show a clear comparison or extreme example
-Use a split scene with contrasting states if applicable
-Include a reference object/figure for scale-based adjectives
-Use universal symbols (↑↓ for tall/short, 🌡️ for hot/cold)
-Emphasize the quality through size, position, or proportion
-For emotions: use simplified facial expressions and body language

Style requirements:
-Bold black outlines (3px weight)
-Flat, solid fill colors
-High contrast shapes
-Centered in frame
-Pure white background
-No gradients or shadows
-1:1 aspect ratio

Return only the prompt, no explanations. Keep it under 100 words.