AI Recipe Automator - Pinterest & WordPress

AI Recipe Pipeline: Gemini 3 Pro + Gemini 2.5 Flash analyze photos & generate SEO articles, Flux AI creates Pinterest visuals, auto-publish pins. Managed via Google Sheets.

Comment

Inspiration

I've always been passionate about automation systems—even before the AI revolution. When I invested $1,000 in a blogging course, I quickly realized the painful truth: creating quality content requires massive manual work. From researching recipes, writing SEO-optimized articles, designing Pinterest graphics, to scheduling posts—it was overwhelming.

As a die-hard Google products fan (especially Google Apps Script!), I thought: "Why not automate the entire workflow?" That's when AI Recipe Automator was born—a project that started as small automation scripts and evolved into a full AI-powered content pipeline.

What it does

AI Recipe Automator transforms a simple food photo into:

  • 📝 SEO-optimized WordPress articles with structured recipe data
  • 🎨 Custom Pinterest graphics generated by AI
  • 📌 Auto-published Pinterest pins with optimized descriptions
  • 📊 Cost tracking for every AI generation

All managed from a single Google Sheets interface—no coding required for daily use.

How I built it

The entire system runs on Google Apps Script (GAS), leveraging:

  • Gemini 3 Pro Preview: Analyzes recipe images, extracts ingredients, generates structured article content with vision capabilities
  • Gemini 2.5 Flash: Fast text generation for Pinterest descriptions and SEO titles
  • Replicate (Flux Kontext): AI image generation to create clean, text-free Pinterest visuals from recipe photos
  • WordPress REST API: Automated article publishing with WP Recipe Maker integration
  • Pabbly Connect: Webhook orchestration for Pinterest API integration
  • Google Slides: Template-based image generation as an alternative to AI

Challenges I faced

  1. Pinterest API limitations: Pinterest doesn't allow direct pin creation via API for personal accounts. I solved this using Pabbly Connect as an intermediary.

  2. Large numeric ID handling: Pinterest IDs are 19+ digits, which JavaScript rounds due to floating-point limitations. Fixed by forcing string conversion before JSON parsing.

  3. Google Drive URL compatibility: External APIs couldn't access Drive files. Implemented automatic URL transformation to direct download links.

  4. Multi-account management: The system now supports multiple Pinterest accounts and WordPress sites from a single spreadsheet.

What I learned

  • The power of combining multiple AI models for different tasks
  • That Google Apps Script is incredibly underrated for building production-ready automation
  • How to architect a modular codebase that's easy to extend

What's next

  • Instagram Reels auto-publishing (already in progress!)
  • TikTok integration
  • AI-powered A/B testing for Pinterest descriptions

Built With

Share this project:

Updates