Source: https://github.com/aipoch/medical-research-skills

When to Use

• Brand marketing campaigns where you need a hero visual that matches brand tone and supports headline/copy overlay.
• Event promotions (e.g., seasonal sales, product launches) requiring a strong focal point and clear layout guidance.
• Website/app banners that must fit specific aspect ratios and maintain readable negative space for text.
• Social media hero creatives where you need fast iteration from vague requirements to a consistent visual direction.
• Pitching multiple design directions to stakeholders (e.g., 2–3 KV concepts with distinct styles and compositions).

Key Features

• Converts vague KV requirements into professional design directions (visual focus, layout suggestions, style details).
• Produces image-generation-ready prompts optimized for KV composition and marketing readability.
• Encourages negative space planning for copy placement and subject prominence for attention capture.
• Supports common KV constraints: scenario-driven style, main color tone, and size/ratio intent.

Dependencies

• Python 3.10+
• Required Python packages:
  • zhipuai
  • requests
• ZhipuAI API access via environment variable:
  • ZHIPUAI_API_KEY (required)

Example Usage

1) Collect KV Requirements

Prepare the following information:

• Project/Brand Name
• Core Slogan/Copy (the “soul” of the KV)
• Usage Scenario (e.g., New Product Launch, Black Friday Sale, Website Banner)
• Style Preference (e.g., 3D render, minimalist, illustration)
• Main Color Tone
• Size/Ratio Requirements (even if the script defaults to one, capture the user’s intent)

2) Analyze the Design Direction

Use the KV Design Analysis Prompts in:
references/design_prompts.md

Process:

1. Open KV Design Analysis Prompts.
2. Fill in the collected information into [KV Design Requirements].
3. Generate:
   • Visual Focus
   • Layout Suggestions
   • Style Details

3) Generate the KV Image Prompt

Use the KV Image Prompt Generation Prompts in:
references/design_prompts.md

Process:

1. Open KV Image Prompt Generation Prompts.
2. Fill in:
   • [Basic Requirements]
   • [Design Proposal] (from Step 2)
3. Generate the final optimized image prompt.

4) Generate the KV Image (Runnable)

1. Ensure the API key is set:

   export ZHIPUAI_API_KEY="YOUR_KEY_HERE"

2. Run the script with the generated prompt:

   python scripts/generate_kv.py "<generated_prompt>"

3. After completion, return the generated image file path to the user.

Implementation Details

• Two-stage prompting workflow

  1. Design analysis stage: transforms requirements into actionable design guidance (focus, layout, style).
  2. Prompt generation stage: converts the design guidance into a production-ready image prompt.

• KV composition constraints

  • Negative space: reserve clean areas for headline/subcopy overlays; avoid overly busy backgrounds.
  • Subject prominence: ensure the primary subject (product/concept) is visually dominant and attention-grabbing.

• Parameters to capture early

  • Scenario drives composition and tone (e.g., “sale” vs. “premium launch”).
  • Style preference constrains rendering approach (3D/illustration/minimal).
  • Main color tone ensures brand consistency.
  • Aspect ratio/size intent informs layout planning even if generation defaults to a preset.