Content Automation Pipeline

Making short-form content every day was eating my mornings. Over an hour, every day, on the same sequence: search Reddit for trending topics, pick one, write a script, hunt for copyright-free footage, record audio, sort everything into folders.

Repetition on a fixed schedule is exactly what machines are for.

Saving the hour was only half the point. I also wanted to learn workflow orchestration properly: how n8n composes services, and what it actually takes to run an unattended AI pipeline you can trust.

So I set a bar. It runs every morning whether or not my laptop is on. It never fails silently. And I can reconstruct any run from its logs.

A schedule trigger fires at 6 AM and drives fifteen nodes: fetch trending Reddit topics, choose one, generate a script with Gemini Flash, clean the output, then hit a validation gate. A valid script flows on to Pexels for stock video, narration generation, Google Drive for storage, a log row in Google Sheets, and a summary email to me. An invalid script takes the failure branch instead: an error email, then a safe stop.

The whole thing runs on Railway using the official n8n template backed by PostgreSQL, so execution history survives redeploys and credentials live in managed environment variables. None of it depends on my hardware.

The pipeline now runs every morning with zero manual steps. Scripts, narration, and stock footage get produced automatically, assets land organised in Google Drive, every execution is tracked in Sheets, and there's a summary email waiting when I wake up. That's roughly two hours back, daily.

The learning mattered as much as the hours. Chaining six APIs, prompt engineering for output a machine has to consume, and designing error handling for a system nobody is watching taught me more about automation than any tutorial could.