Instagram Slides

Turn blog posts or content into Instagram carousel slideshows.

Pipeline

1. Extract — fetch blog post content via web_fetch
2. Plan — write a plan.json with slide content and image prompts
3. Generate Backgrounds — AI image generation via Fal API
4. Text Overlay — Pillow composites text onto backgrounds with brand styling
5. Output — numbered slide images + caption.txt

Usage

# Full pipeline
python3 {baseDir}/scripts/generate.py \
  --url "https://blog.example.com/post" \
  --slides 8 \
  --output ~/Desktop/slides

# Plan only (review before generating)
python3 {baseDir}/scripts/generate.py \
  --url "https://blog.example.com/post" \
  --slides 8 \
  --plan-only \
  --output ~/Desktop/slides

# Generate from existing plan
python3 {baseDir}/scripts/generate.py \
  --plan-file ~/Desktop/slides/plan.json \
  --output ~/Desktop/slides

Plan JSON Schema

{
  "angle": "how-to | listicle | hot-take | story-arc | myth-busting",
  "style_prefix": "Shared image generation prompt prefix (~30 words)",
  "slides": [
    {
      "headline": "Short headline (≤8 words)",
      "body": "Supporting text (≤20 words)",
      "bg_prompt": "Background image description (NO text in image)",
      "is_title": true,
      "is_cta": false
    }
  ],
  "caption": "Full Instagram caption with hook, value, CTA, hashtags"
}

Image Style Tips

Good:

• Bright, warm, natural lighting
• Clean, airy, modern aesthetic
• Contextual imagery illustrating each slide's topic
• Editorial photography feel

Avoid:

• Baked-in text/words in generated images
• Dark, moody backgrounds (unless that's your brand)
• Abstract meaningless gradients

Dependencies

• Python 3 with Pillow, requests: pip3 install Pillow requests
• Fal API key at ~/.config/fal/api_key
• Brand fonts in {baseDir}/fonts/
• OpenRouter API key for auto-planning (optional)

Costs

• ~$0.15/image via Fal
• ~$1.20 for 8 slides

Tips

• First slide = strong hook/title card
• Last slide = CTA with link
• Keep text SHORT — Instagram is visual-first
• Raw backgrounds saved as raw_NN.png — re-composite without regenerating images