Case Study
YouTube → Content
A fully automated Telegram bot that turns any YouTube video into ready-to-post social media content. No manual summarizing, no watching videos just one sharp, engaging post per video.
Overview
What is YouTube → Content?
In plain English
You send a YouTube link to a Telegram bot. The bot fetches the full transcript, reads it with AI, finds the single most interesting or surprising moment in the entire video, and writes content around that one moment. You copy and post. Done.
Most content creators and marketers waste hours watching YouTube videos just to extract one usable idea for a post. This tool eliminates that entirely. Drop a YouTube link into Telegram, and within seconds you receive content not bullet points, but one specific moment from the video that will make someone stop scrolling.
Pipeline Breakdown
How it works
The backend handles the extraction, processing, and formatting automatically without human intervention.
Step 1: Trigger
User sends a YouTube URL to a dedicated Telegram bot, initiating the webhook listener.
Step 2: Transcript Retrieval
n8n intercepts the link and queries the Supadata API to instantly fetch the video transcript text.
Step 3: AI Core Processing
The transcript is sent to Google Gemini Flash with strict rules to find a single hook and build structured output.
Step 4: Output Delivery
The formatted social media content is sent directly back to the user inside the Telegram chat window.
Key Capabilities
Features and Integration
Engineered to provide a zero-friction experience for immediate social sharing.
Ignores generic summaries. Focuses strictly on identifying unique statistics, counter-intuitive arguments, or key quotes.
Operates entirely inside Telegram. No extra logins, accounts, or dashboard panels to navigate.
Runs the entire pipeline from link extraction to content delivery in under 30 seconds.
Key Engineering Challenge
The prompt was the hardest part to build.
Early versions of the AI output were generic summaries the kind of content nobody stops to read. The post sounded like a Wikipedia entry, not something a human would share.
The Problem
Generic summaries that did not work
The early AI outputs were boring, generic summaries that wouldn't make anyone stop scrolling on social media. They read like robotic Wikipedia descriptions rather than engaging, human-written posts.
What I Built
One strict rule transformed everything
I rewrote the Gemini prompt with one non-negotiable rule: do not summarize the video. Instead, find one specific moment, data point, or quote, and write the entire post around that single element.
- ✓ Specific, shareable moments rather than general summaries
- ✓ Under 30-second pipeline execution
- ✓ Completely hands-off trigger and delivery inside Telegram
Technology
Platform Technologies
A lightweight automated workflow stack combining specialized APIs and serverless execution.
Visual workflow automation tool used to connect webhooks, logic blocks, and external APIs.
State-of-the-art AI model responsible for processing transcripts and generating structured copy.
Specialized API used to retrieve accurate, raw YouTube transcripts via video URL parameters.
Handles real-time communication between the end user and the backend automation workflow.
Interface Walkthrough
n8n Workflow Screenshot
Click the screenshot to open a full-size lightbox.
Need a workflow like this?
If you have a repetitive content or data task that should be automated, I build these pipelines end to end.