BlogAI

Turn videos into Words

Comment 1

Inspiration

"What if AI could watch these videos and generate blog posts instantly?"

The idea for this project was born out of a desire to bridge the gap between video content and written blogs — especially as the internet grows more saturated with diverse formats. While YouTube is rich with information, many users prefer or need content in a written form: blogs, articles, and summaries.

But here's the twist — I wanted the tool to generate context-aware, real-time relevant blogs, not just stale summaries.

This led me to leverage Google's Agent Development Kit (ADK) to fetch current trends and contextual insights, ensuring the generated blogs are not only derived from the YouTube video but also enriched with timely, relevant information pulled directly via Google Search and Trends APIs.

What it does

  • Transforms any YouTube video into a ready-to-publish, AI-generated blog.
  • Allows customization of:
    • Length: Short, Medium, Long.
    • Tone: Conversational, Professional, Neutral.
  • Enriches content using Google's ADK agents to fetch current trends, popular subtopics, and real-time details related to the video's theme.
  • Fully hosted and deployed on Google Cloud Run, ensuring scalability, reliability, and seamless updates.
  • Offers a clean dashboard for users to save, view, and manage their generated blogs.
  • Implements custom authentication.
  • Upcoming features like "Publish to Medium" and Image Generation planned.

How we built it

Tech Highlights:

  1. Frontend: • Built using Next.js (App Router), with modern UI via Tailwind CSS and shadcn/ui. • Responsive with Dark/Light theme switching.

  2. Authentication: • Powered by NextAuth.js with Google OAuth integration.

  3. Backend & AI Workflow: • Core AI flow built using Google Genkit to handle prompt orchestration, AI inference, and output formatting. • ADK agents:

    • Pull Google Trends data related to the video.
    • Fetch supporting articles, subtopics, and breaking news connected to the video theme.
    • Feed this into the Genkit AI pipeline, enriching the final generated blog. • The agent responsible for fetching contextual, up-to-date information (like trends, related topics, fresh content summaries) was built with Google's ADK (Agent Development Kit). • Deployed as a containerized microservice on Google Cloud Run.

  4. Database & Storage: • PostgreSQL, managed locally via Docker during dev, and Google Cloud SQL in production. • Prisma ORM used for type-safe schema and queries.

  5. Deployment: • Entire application containerized via Docker. • Deployed on Google Cloud Run for serverless scalability. • Managed environment variables, secrets, and scaling via Google Cloud Build + Artifact Registry.

Challenges we ran into

  • Schema Management: Handling Prisma schema changes and migrations between local Docker DB and production Cloud SQL instance.

  • Deployment Complexity: Configuring CI/CD for seamless builds and auto-deployments to Cloud Run.

  • Agent Integration (ADK): Designing and training custom ADK agents to query Google Search, Trends, and Knowledge Graph APIs efficiently.

  • Combining AI Flows (Genkit + ADK): Coordinating Genkit AI pipelines with real-time ADK agent outputs — ensuring consistency and performance without overloading generation time.

  • Authentication Management: Balancing Google OAuth flow with optional email/password login, while maintaining secure access to user-specific content.

  • Resource Optimization: Making the pipeline cost-effective on Cloud Run while handling scalable requests.

  • Content Quality Assurance: Ensuring the AI-generated blogs remained coherent, SEO-friendly, and context-aware, especially when ADK trends suggested rapidly changing or volatile topics.

Accomplishments that we're proud of

  • 🎉 Built a fully functional, production-ready website capable of:

    • Turning YouTube videos into rich, customized blog posts.
    • Enriching content with real-time, AI-curated trends and details via Google's ADK.
    • Smooth AI flow management with Google Genkit.
    • Fully deployed on Google Cloud Run with auto-scaling.
    • Secure and scalable user authentication.
    • A clean, user-friendly dashboard for post management.

  • 🚀 Successfully designed, built, and deployed a multi-agent, AI-driven system leveraging modern Google Cloud and AI tooling.

Next Plans

  • 🔜 Publish to Medium: Let users directly push generated blogs to their Medium accounts via the Medium API.

  • 🎨 AI-Based Image Generation: Auto-generate cover images for each blog using Google Imagen or DALL·E integration.

  • 🌍 Multi-language Blog Support: Enable users to generate blogs in multiple languages.

  • 📝 Inline Blog Editor: Allow users to tweak the AI-generated post before saving or publishing.

  • 📊 SEO Analyzer Integration: Provide real-time SEO score and suggestions for the generated blogs.

  • 🤖 Custom ADK Agents for Niche Topics: Build specialized ADK agents for technology, health, finance — delivering hyper-personalized content.

  • 🧩 Chrome Extension: One-click blog generation directly from YouTube video pages.

Built With

Share this project:

Updates