← All work
live · 2025

YouTube Synthesizer

Paste a URL, get a transcript, AI summary, sentiment, and Google Trends overlay.

YouTube Synthesizer — Portfolio Context & Design Principles

The YouTube Synthesizer is a curated AI-powered video summarization tool built as part of the franklydata.com portfolio, a playground for editorial data projects inspired by pudding.cool.


🎯 Product Context

Lengthy YouTube videos contain valuable insights but require significant time to consume. The YouTube Synthesizer bridges this gap by automatically converting any video URL into a comprehensive, beautifully structured HTML report. It parses the video metadata, extracts the transcript, detects sentiment, and designs a clean, readable visual digest.

Core Features

  • Instant Summarization: Transforms videos into fully formatted, readable summaries in seconds.
  • Sentiment & Mood Extraction: Employs AI to score the emotional profile of the speaker's transcript.
  • Visual Transcript Explorer: Synchronizes timestamps with sections to allow fast scanning.
  • Developer Test Panel: An inline developer console built directly into the UI to allow direct webhook payload testing and debugging.

🏗️ Architecture & Integrations

The YouTube Synthesizer is built on a highly modular, decoupled architecture, delegating its complex AI workflows to an external MindStudio agent system via webhook.

graph TD
    User([User URL Input]) -->|POST /api/summarize| Next[Next.js / Express Backend]
    Next -->|Call Webhook| MS[MindStudio Orchestrator]
    MS -->|Fetch Video| YT[YouTube Data API]
    MS -->|Transcribe & Analyze| Claude[Claude Sonnet 4]
    Claude -->|Return HTML Synthesis| MS
    MS -->|JSON Response| Next
    Next -->|JSON Response| Client([Client Webpage])
    Client -->|Direct HTML Injection| DOM[Render Summary View]

Technical Workflow

  1. User Action: The user inputs a YouTube URL (supporting formats like watch?v=, youtu.be, /embed/, or query parameters) and submits.
  2. Backend Delegation: The Express/Next.js backend receives the request at /api/summarize and handles validation, rate limiting, and request headers.
  3. MindStudio Pipeline: The backend triggers a POST request to the MindStudio webhook. MindStudio orchestrates the multi-agent pipeline:
    • Fetches video metadata via YouTube Data API.
    • Extracts the video transcript.
    • Summarizes and formats content using Claude Sonnet.
    • Generates a customized, styled HTML structure representing the summary.
  4. Client-Side Rendering: The response is returned to the client in a JSON field (site_html), which is immediately injected into the result container within a safe viewport.

🎨 Design Principles & Aesthetics

The design is engineered to evoke the experience of an editorial data story:

  • Rich Dark Aesthetics: Deep charcoal and pitch-black backgrounds paired with neon-violet gradients create a sophisticated, tech-focused interface.
  • Aesthetic Typography: Bold sans-serif header typography paired with clean, readable body fonts optimized for scanning long-form summaries.
  • Developer-Friendly Diagnostics: The UI incorporates a collapsible console-style panel at the bottom, allowing real-time testing of different API payloads (valid links, invalid URLs, expired cookies) to make developer collaboration fluid.
  • No-Persistence Privacy: To guarantee absolute user privacy, transcripts and summaries are processed entirely in memory. The tool does not store user URLs or generated reports in any database.

🛠️ Technology Stack

  • Frontend: Vanilla HTML/JS, tailwindcss.
  • Backend: Node.js, Express, TypeScript, Next.js.
  • Orchestration: MindStudio Webhook API (V2).
  • AI Processing: Claude (Anthropic).
  • Hosting: Docker containerized, deployed on Google Cloud Run.